Authors: Shaoting Tan, Ning Liu, Yuntao Du, Shuyue Wei, Wu Shuai, Qian Li, Yanyu Xu, Wei Zhang, Lizhen Cui, Haitao Yuan
Abstract: Sequential diagnosis requires balancing diagnostic accuracy against resource costs through iterative information gathering. Existing Large Language Model (LLM) approaches exhibit a critical knowledge-reasoning gap: despite encoding extensive medical knowledge, they struggle to reason systematically under cost constraints, often resorting to excessive testing. We propose GraphDx, a knowledge-enhanced framework with two core innovations. First, we design an automated pipeline that leverages LLMs to construct Medical Diagnosis Knowledge Graphs (MDKGs) with quantized typicality, action-centric topology, and dual-objective attributes for both diagnostic relevance and cost-sensitivity. Second, we introduce three collaborative agents (Perception, Reasoning, and Decision) where the Perception and Decision Agents handle language understanding and generation, while the Reasoning Agent performs deterministic evidence scoring and cost-aware planning on the MDKG. Experiments on MedQA and MIMIC-IV across three LLM backbones (DeepSeek-V3, Kimi-k2, Llama-3.3) show that GraphDx improves diagnostic success rates from 50--68% to 79--93% while reducing test costs by 20--54%, providing a robust, economical, and interpretable solution for automated clinical diagnosis.
Authors: Su Lan, Xuefei Yin, Yanming Zhu, Alan Wee-Chung Liew
Abstract: Causal and intervention-based question answering is fundamental to advancing large language models (LLMs) toward reasoning beyond surface-level correlations and understanding underlying causal mechanisms. However, existing LLM-based methods often rely on implicit language-level reasoning, resulting in opaque causal assumptions, unverifiable reasoning paths, and fragile predictions under complex interventions, particularly in context-free settings. In this paper, we propose an explicit and auditable causal reasoning framework for context-free intervention-based question answering. Our method formulates causal inference as structured reasoning over an explicit causal graph through four modular stages, rather than implicit end-to-end prediction. A key innovation is a target-aware causal graph construction strategy that treats the target variable as a core constraint during graph expansion, effectively suppressing irrelevant variables, spurious causal relations, and reasoning noise. We further introduce a path-level causal evidence aggregation mechanism that combines multiple causal paths while modeling both reinforcing and counteracting effects, enabling robust decision-making beyond single-chain reasoning. Extensive experiments on three benchmarks demonstrate that our framework consistently outperforms existing LLM-based methods while providing interpretable and auditable causal reasoning traces.
Authors: actAVA AI, :, Haolin Chen, Leon Qi, Steve Brown, Deon Metelski, Tao Xia, Joonyul Lee, Qixuan Wang, Kevin Riley, Frank Wang, Weiran Yao
Abstract: Healthcare spans high-stakes communication, expert reasoning, and workflow execution, yet specialized LLMs that cover these use cases together remain limited. A healthcare model must handle patient consultation, clinical reasoning over text and images, interactive diagnosis, and electronic health record (EHR) tool use. These capabilities fail in different ways, and a narrow update for one task can degrade another. We present Cura 1T, a healthcare-specialized LLM trained through a human-gated self-evolution loop. In each evolution round, a training agent plans a target capability, trains the model, evaluates benchmark trajectories, and refines the data mixture from observed failures. This data-centered loop improves the model through targeted synthetic and curated examples rather than a single generic medical-data update. Across the healthcare evaluation suite, Cura 1T ranks at or near the top among frontier baselines, while remaining competitive on out-of-domain reasoning and agentic benchmarks.
Authors: Raunak B Sinha
Abstract: Desktop voice assistants are still dominated by cloud pipelines that ship raw audio off the machine and expose a fixed set of skills. We describe AnovaX, a small local-first assistant that runs entirely on the user's computer and treats the desktop itself as its action surface. A single Python process wires together a wake-word gate, a speech pipeline, an LLM planner (Gemini) that emits a JSON plan of tool calls, a whitelist-and-denylist safety layer, a multi-agent orchestrator that translates each plan into typed child agents on a bounded thread pool, and an adaptive recovery loop that takes over whenever a core step fails. Every tool corresponds to a specialized agent class (AppAgent, TypingAgent, BrowserAgent and six others) with its own timeout, retry policy, and shared-resource locks. A recursive MetaAgent lets the planner delegate a sub-goal back to itself, capped at two levels of nesting. The recovery loop uses a compact ReAct-style prompt and hides Gemini's latency behind speculative execution of read-only tools. A companion Flask server exposes a phone-friendly remote over the local WiFi, mirrors every agent lifecycle event to the phone in real time, and streams the laptop's screen back over MJPEG so the user can watch remote commands land as they run. The point of the project is less to compete with Siri or Alexa than to show that a legible, few-thousand-line assistant is enough to open apps, type into them, run searches, coordinate concurrent actions, recover from single-step failures, and be driven entirely from a phone in another room -- without the LLM ever touching the keyboard.
Authors: Chih-Hsuan Yang, Jingyan Jiang, Vikram Vasudevan, Cheng-Hau Yang, Huihuo Zheng, Le Chen, Eliu A. Huerta, Venkatram Vishwanath, Ian T. Foster, Rajeev Thakur
Abstract: Many math- and science-oriented agent systems use hierarchical designs with specialized reviewer roles, assuming that a dedicated review stage should help turn wrong candidates into correct ones. We test this assumption on 4,181 verifier-grounded Omni-MATH problems using matched gpt-oss-120b actors. Collaboration adds little on the easiest tiers, but from tier 4 onward the gains open sharply; in this harder regime, broadcast-style peer discussion reaches higher final accuracy than a planner-executor-reviewer pipeline (PER). We ask whether this gap is explained by reviewer quality or by whether critique changes the next answer the protocol carries forward. It is not explained by reviewer precision alone: PER's reviewer is more precise than broadcast's (0.861 vs. 0.644), yet evaluator-verified useful critique is much less likely to change the next candidate and produces lower reviewer-guided repair. These results show that reviewer detection quality and critique uptake are empirically separable. Within matched PER interventions, forcing explicit acknowledgment lowers final accuracy, while embedding reviewer guidance directly in the solver's working context partially improves follow-through without closing the gap. Overall, reviewer-centric evaluation can overstate system quality: a protocol may spot errors well yet still fail to solve more problems if it does not act on those critiques.
Authors: Yoonhwa Jung, Junryu Fu, Mani Golparvar-Fard
Abstract: We introduce DrawingVQA, the first benchmark designed to evaluate multimodal large language models (MLLMs) on real-world construction drawings -- a core media in architecture, civil, and many other engineering practices. Unlike natural images or schematic floor plans, construction drawings fuse abstract geometry, symbolic notation, tabular data, annotations, and domain-specific text, forming a uniquely complex visual-textual domain core to engineering workflows. DrawingVQA bridges this gap with 33 "Issued for Construction" drawings and 92 expertly curated question-answer pairs, spanning three reasoning depths: perceptual understanding, contextual interpretation, and domain-expert reasoning. To evaluate model capabilities, we present a dual categorization framework to jointly analyze performance across seven construction-engineering and four MLLM capability dimensions -- the first to explicitly map engineering workflows to AI reasoning competencies. Evaluations of state-of-the-art MLLMs reveal a substantial gap between model and expert performance, particularly at higher reasoning depths. This benchmark lays a foundation for domain-specialized multimodal reasoning to allow for advancement on integration of AI-driven understanding and real-world engineering workflows.
Authors: Sergey Rodionov
Abstract: Our previous ARC-AGI-3 agent bundled executable world modeling, scheduled simplification, and exact replay verification, leaving unclear which idea accounted for its performance. We address this attribution question with four nested Codex-based agents: a textual baseline; a flexible-interface executable world model without replay verification; the same executable model with scheduled simplification; and a fixed-interface verification treatment that retains simplification and requires exact reproduction of recorded observations. The main study evaluates all four agents with gpt-5.4 and gpt-5.5 at high and xhigh reasoning effort on the public ARC-AGI-3 games. Exploratory follow-ups evaluate the textual and verification variants with gpt-5.6-sol at xhigh and max. The most robust result is that every agent variant improves with a stronger model and with greater reasoning effort. Within each model-effort setting, differences among variants are smaller than anticipated, while the effects of individual components vary across settings. Requiring a persistent executable deliverable is not universally beneficial: the textual variant outperforms the flexible-interface executable variant in both gpt-5.5 settings. Simplification improves performance in three of the four model-effort settings, with the weakest setting as the only exception. The complete verification treatment ranks first in all four settings, although it uses substantially more resources. In the gpt-5.6-sol follow-up, the verification variant fully solves every public game at both reasoning efforts, achieves about 99% RHAE, and uses fewer than half the total actions of the human baseline. Because the model postdates these games and held-out performance remains untested, this result should be interpreted as saturation of the public set only.
Authors: Shanhong Liu, Pai Chet Ng, De Wen Soh, Malika Meghjani, Konstantinos N. Plataniotis
Abstract: Internet memes intertwine visual cues, textual content, and cultural context, making them particularly challenging to interpret in scenarios where humor, sarcasm, and harmful intent coexist. These complexities highlight the need for explainable meme understanding systems that can provide reliable and structured reasoning to support both accurate classification and human interpretability. However, existing multimodal classifiers either overlook these interdependencies or provide only limited interpretability. In this paper, we introduce MAR-12, a novel framework that leverages Vision Language Models (VLMs) for meme detection and understanding in settings where humorous and hateful elements may coexist. The framework first interprets each meme through twelve structured perspectives derived from humor and hate theories. It then applies a role-aware soft-gated attention mechanism to learn how much each perspective should contribute, followed by a prototype-based classifier for the final prediction. Finally, explanations are synthesized using both perspective-specific reasoning and learned attention weights, ensuring transparent and context-grounded justifications. We evaluate MAR-12 on the PrideMM and Memotion datasets, where it achieves up to 80.3% accuracy for humor detection and 75.9% accuracy for hate detection, outperforming state-of-the-art approaches. Furthermore, both human and GPT-4-based evaluations confirm that MAR-12 produces coherent and persuasive explanations, particularly for memes in which humorous and harmful cues co-occur.
Authors: Eduardo C. Garrido-Merch\'an
Abstract: A trained deep reinforcement learning policy is a black box, and we ask whether it can be made explainable by rewriting it as an executable logic program that reproduces its behaviour and that a person can read, a logic engine can run, and an optimizer can edit. We present a three-stage post-hoc transformation that extracts a frozen proximal policy optimization teacher, induces an ordered rule list from its decisions in the manner of classical relational learning, and emits the result as a Prolog program whose every decision is executed by an off-the-shelf logic engine; a subsequent expansion stage edits the rule base and accepts an edit only when policy evaluation certifies a return increase. We prove four guarantees. A return-loss bound makes the distilled program a machine-checkable certificate in a finite Markov decision process, and the expansion loop improves monotonically and terminates. For the continuous-observation setting we answer whether the conversion is possible at all: the propositional threshold instantiation converts the network to arbitrary fidelity as the resolution B grows, with disagreement O(1/B) and a return gap that closes at the same rate, and a matching lower bound shows the cost is exponential in the observation dimension for an oblique decision boundary. Empirically, on a two-room key-and-door task with 16,944 reachable states the expanded Prolog program attains exact optimal return in every seed and, in a budget-capped regime, exceeds the stochastic teacher on exact return in ten of ten seeds. On three continuous-control tasks the emitted program substitutes the network, matching the neural teacher within noise on Acrobot with eleven clauses and recovering about 97% of its return on CartPole, while on the finer-control LunarLander it recovers only partially, exactly the ceiling the exponential lower bound predicts.
Authors: Michael Papademas, Xenia Ziouvelou, Kostas Karpouzis, Vangelis Karkaletsis
Abstract: As artificial intelligence (AI) systems increasingly impact society, ensuring their ethical and trustworthy deployment has become a global priority. While a myriad of high-level ethical guidelines have emerged, criticism persists that these frameworks remain abstract and lack concrete mechanisms for implementation. This paper conducts a critical analysis of tools and trust mark frameworks intended to operationalize trustworthy AI (TAI), drawing on a comprehensive dataset from the OECD. Through empirical mapping and descriptive comparative analysis, we identify significant asymmetries in ethical focus, lifecycle coverage, stakeholder targeting, and tool typology. Our findings show a strong emphasis on fairness, transparency, and robustness, with comparatively little attention paid to explainability, digital security, and environmental sustainability. Moreover, most tools and certifications concentrate on post-development stages, with limited guidance for early design or data collection phases. Educational initiatives and policy engagement are notably underdeveloped, suggesting that current TAI efforts are dominated by technical and procedural measures within industry contexts. We argue that bridging the persistent chasm between AI principles and practice requires expanding ethical objectives, embedding ethics across the AI lifecycle, and fostering broader multi-stakeholder participation. This study provides both a diagnosis of existing implementation gaps and actionable recommendations for advancing more holistic, inclusive, and enforceable AI governance
Authors: J. N. Hooker
Abstract: Logic and optimization can, in combination, make valuable contributions to rule-based AI. Logic is the obvious medium for encoding a rule base and drawing inferences from it, while optimization provides a powerful technology for computing inferences. Their combination has taken on new relevance amid a growing concern for transparency in AI. which is important for reproducibility, explainability, trustworthiness, and fairness. Rule-based AI provides a natural solution to transparency that is becoming increasingly practical due to today's highly advanced optimization methods. This article surveys several areas of logic-optimization partnership, including probabilistic logic, Bayesian logic, belief logics and Dempster-Shafer theory, nonmonotonic (default) logic, many-valued logics, and inference of logical formulas from noisy data based on Boolean regression. It shows how to compute projections, the fundamental problem of both logic and optimization, using decision diagrams and logic-based Benders decomposition. It describes the use of postoptimality analysis to explain how conclusions are reached, further enhancing transparency, as well as the role of optimization in answer set programming modulo theories. The paper concludes by suggesting possible future research directions.
Authors: Xue Yu, Bo Yuan, Pengshuai Yang, Kailin Zhao, Hong Hu, Junlan Feng
Abstract: Mobile graphical user interface (GUI) agents have demonstrated remarkable capabilities in automating complex tasks, yet they introduce critical safety risks where a single erroneous action can lead to irreversible consequences. Existing safety mechanisms are primarily reactive, lacking the ability to assess risks before execution. In this paper, we introduce SeerGuard, a consequence-aware safety framework designed to mitigate these risks through pre-execution instruction-level screening and action-level risk assessment. Specifically, the action-level assessment analyzes agent-proposed actions within current GUI states, anticipating likely outcomes to identify risks before they are executed. To enable these capabilities, we construct a unified safety-augmented world model (SAWM) via multi-task learning, integrating semantic next-state prediction with safety risk assessment. Extensive experiments demonstrate that SeerGuard generalizes effectively across diverse mobile GUI agents. On Qwen3-VL-8B-Instruct, it increases the safety-utility score from $0.191$ to $0.596$ at $\omega=0.8$ and reduces the risk-cost score from $0.347$ to $0.130$ at $\alpha=0.8$. Further analyses on our SAWM validate the effectiveness of the instruction-level screening, alongside the capability of action risk assessment and next-state prediction.
Authors: Xu Hou, Meiyu Liang, Wei Huang, Yawen Li, Zhe Xue, Wu Liu, Guanhua Ye, Lei Shi, Kangkang Lu
Abstract: Multimodal Knowledge Graph Completion (MKGC) requires inferring missing entities from structural, textual, and visual cues. Existing diffusion-based MKGC methods usually denoise directly on raw multimodal features. Such a design forces the denoiser to simultaneously perform relation-dependent cue selection, cross-modal semantic alignment, and structure-aware entity generation, which introduces noisy and semantically inconsistent conditions for diffusion and consequently leads to suboptimal completion performance. To address this limitation, we propose MGDT: MLLM-Guided Diffusion Transformer with Relation-Adaptive Mixture-of-Experts (MGDT), a novel MKGC framework built on an align-then-diffuse paradigm. MGDT first employs a Relation-Adaptive Semantic Routing Mixture-of-Experts (RASR-MoE) module to select relation-relevant multimodal semantic transformation paths and suppress irrelevant modality interference. MGDT then uses a frozen Multimodal Large Language Model (MLLM) as a semantic anchor to align the routed multimodal representations into a unified latent space and reduce cross-modal semantic heterogeneity. Finally, a Knowledge Graph Diffusion Transformer (KGDT) performs graph-conditioned denoising generation in the aligned space to produce the missing entity representation. Experiments on three benchmark datasets show that MGDT consistently outperforms strong baselines.
Authors: Aritro De (The University of Texas at Austin), Juliana Felkner (The University of Texas at Austin)
Abstract: LEED v4.1 BD+C certification remains a document-intensive process that requires reviewers to read hundreds of pages of project evidence and apply credit-specific threshold logic by hand. This paper investigates whether small, locally deployed language models can perform meaningful screening of LEED documentation and how deterministic symbolic components should share that work. A neuro-symbolic pipeline is introduced that aligns project PDFs to LEED credit sections, retrieves evidence with credit-aware keyword signatures, verifies compliance with a locally hosted 4-billion-parameter language model, and applies a LEED-specific numeric checker to quantitative thresholds. Experiments on four university buildings (484 PDFs, 153 credit-level decisions) show that a 4-billion-parameter model (gemma3:4b) is the strongest text-only core verifier, achieving 67.3% accuracy and outperforming a larger 8-billion-parameter model (llama3.1:8b) in this task. The deterministic numeric checker corrects arithmetic errors on key quantitative credits, moving EA-p2 from 50% to 100% accuracy and improving several other credits when required values are reliably extracted. At the same time, the full neuro-symbolic configuration achieves 61.6% overall accuracy, trailing the best text-only baseline due to extraction failures and conservative behavior on qualitative categories. Systematic ablations show that adding low-resolution drawing images (150-300 dpi) consistently reduces accuracy, and that prompt effectiveness depends on the building's ground-truth PASS rate: rubric prompts perform best on documentation-rich projects, while chain-of-thought prompts perform best on documentation-lean projects. Within the specific scope of LEED v4.1 BD+C compliance verification over raw project documentation, this pipeline and its baselines provide an initial reproducible reference point for both accuracy and failure modes.
Authors: Shuaiyu Zhou, Fengpeng Yue, Zengjie Hu, Yuanzhe Shen, Chenyang Zhang, feng hong, Cao Liu, Ke Zeng
Abstract: While LLM agents demonstrate strong reasoning abilities in compact and well-defined scenarios, they struggle to maintain robustness and effectiveness when faced with large-scale, diverse, and dynamic real-world environments that demand seamless tool integration. To address this gap, we introduce ToolVerse, a comprehensive framework that scales up agentic RL environments and enables agents to perform complex long-horizon reasoning in Tool-Integrated Reasoning (TIR) tasks. First, ToolVerse automatically builds the massive executable agent training environments from nearly 400 real-world Model Context Protocols (MCPs) that contain about 4500 tools. Second, we propose a task design strategy based on a tool dependency graph, utilizing Dynamic Unlocking Sampling Algorithm to generate long-horizon tasks, and produce GUST (Graph Unlocking Sampling Tasks) dataset. Third, to alleviate the credit assigment problem in long-horizon agentic RL, we propose a fine-grained Turn-Aware Relative Advantage algorithm. We conduct extensive Agentic RL training using ToolVerse and evaluate our framework on serveral agentic benchmarks. Experimental results demonstrate that our framework significantly strengthens LLMs' capabilities in long-horizon tool use, achieving a marked performance boost and showcasing robust reasoning within dynamic environments.
Authors: Jiahao Zhao, Junyi Liu, Lifeng Xu, Nan Xu, Qingli Wang, Qingxiao Li, Tianle Chen, Xiaoyu Wu, Yawen Zheng, Zikai Wang, Guanming Liu, Hequn Zhou, Jingyi Wang, Jingyuan Shu, Keqi Wang, Li He, Songyang Diao, Wenhui Xu, Xinyu Ren, Yaqin Fan, Yujin Zhou, Zhanao Yao
Abstract: We present S1-Omni, a unified multimodal reasoning model for scientific understanding, prediction, and generation. AI for Science (AI4S) has advanced significantly through domain-specific models, tool-augmented LLMs, and scientific language models. However, model capabilities remain highly fragmented, limiting the joint modeling of heterogeneous data, scientific laws, and expert knowledge. S1-Omni addresses this gap by consolidating these capabilities into a single, coherent scientific reasoning model. The architecture of S1-Omni is built upon three core components: unified representation of scientific data, natural-world knowledge alignment, and decoding for domain-specific tasks. First, S1-Omni maps natural-language instructions and scientific objects, including CIF, SMILES, protein sequences, spectra, and scientific images, into a shared representation space. Second, it incorporates scientific laws and expert knowledge into data construction and training, enabling the model to reason from scientific evidence. Third, it performs task-specific decoding to support a broad range of applications, including property prediction, spectrum-to-molecular generation, protein site and structure prediction, and scientific image generation and editing. S1-Omni is trained on S1-Omni-Corpus, which covers 200 scientific tasks and contains millions of reasoning samples, and is evaluated on over 60 scientific benchmarks. It outperforms GPT-5.5 and Gemini-3.1-Pro on most benchmarks and matches or surpasses domain-specific models on several benchmarks. Overall, S1-Omni provides a practical path toward unified scientific modeling.
Authors: Lujia Zhang, Xingzhou Chen, Hongwei Feng
Abstract: Large language model (LLM) agents are increasingly used for complex information-extraction tasks, yet it remains unclear whether agentic components such as reflection and memory lead to observable and controllable improvements over fixed LLM workflows. We study this question through conference-paper dataset extraction, where a system must identify datasets mentioned in scholarly PDFs and produce structured records. We compare a fixed workflow baseline with reflective agent variants and specify an optimized agent condition (S2) that extends the same task with richer PDF tools and dynamic tool selection. Our evaluation emphasizes process-level behavior--including tool execution, retries, reflection, memory use, runtime, and failure recovery--while treating extraction coverage and field completeness as secondary outcome measures. The paper characterizes when agentic mechanisms change system behavior, whether these changes improve task completion, and how the observed failure modes motivate an optimized agent design under the same evaluation harness.
Authors: Hui Yang, Jiaoyan Chen, Yiping Song, Renate Schmidt, Wen Zhang
Abstract: OWL ontologies provide a formal knowledge representation framework that enables semantic reasoning, and have been widely adopted across domains such as healthcare and bioinformatics. In practice, however, real-world ontologies are often incomplete, which pose challenges for reasoning. In this work, we focus on a fundamental subsumption reasoning problem: given an incomplete ontology and a candidate (non-entailed) subsumption, determine whether the subsumption is semantically plausible and, if so, providing a logically sound explanation containing potential missing axioms. This task unifies subsumption verification with ontology abduction, and generalizes the latter by removing the need for a predefined candidate set of missing axioms. To address this subsumption reasoning problem, we propose NeurOWL, an end-to-end neuro-symbolic framework that jointly performs verification and abduction, leveraging both formally defined semantics and textual semantics through Large Language Models and ontology embeddings. We evaluate NeurOWL on real-world ontologies across multiple domains, demonstrating strong and robust performance across different domains.
Authors: Ming Chen, Pranav Pai
Abstract: Geospatial datasets support applications from urban planning to climate modeling, yet consistent assessment of FAIR compliance is difficult. Existing evaluators use different rubrics and evidence sources and may fail on JavaScript-rendered pages or repository-specific identifiers. For 50 datasets from 10 repositories, the standard deviation of normalized scores across available tools averages 15.0 percentage points and reaches 30.3 for one dataset. Because these outputs are not equivalent measurements, we use them to characterize disagreement and failure modes, not comparative accuracy. We present AgentFAIR, a multi-agent framework combining structured metadata extraction with 13 sub-principle-specific LLM evaluators. Each produces a 0-3 maturity score, cited evidence, and recommendations; a critic checks evidence and consistency and can request targeted re-evaluation. Mean Findability, Accessibility, Interoperability, and Reusability scores are 79.7%, 70.4%, 45.3%, and 72.0%. Rank correlations with four baseline tools range from 0.31 to 0.61; the FAIR-enough comparison is not statistically significant. On a 10-dataset repeated-run subset, sub-principle agreement averages 89% (standard deviation: 3 percentage points), versus 71% without the critic. A preliminary 15-dataset expert study yields Fleiss' kappa of 0.71 and 82% alignment with expert consensus. API cost is approximately USD 0.054 per dataset. These results support auditability and feasibility, while the limited benchmark, incomplete ablations, and single-model-family validation constrain claims about accuracy and generalization.
Authors: Zhendong Li, Lei Sun, Ruibo Ming, He Zhang, Danda Pani Paudel, Luc Van Gool, Jinjin Gu
Abstract: Workflow generation in visual creation systems such as ComfyUI demands not only syntactic accuracy but also expert-level reasoning over modular compositions. Existing large language model (LLM) approaches often treat this as a direct text-to-JSON generation task, struggling with structural brittleness and lacking the experiential knowledge required for effective design. We argue that successful workflow generation requires modeling knowledge itself, including its structure, hierarchy, and reasoning dynamics. To this end, we propose a knowledge-centric framework that learns to invert, inject, and infer with knowledge across multiple abstraction levels. We first perform knowledge inversion to distill hierarchical representations, ranging from full pseudo-codes and skeletons to high-level strategies, from large collections of real-world workflows. We then conduct knowledge injection through supervised fine-tuning, teaching the model to reason from task descriptions to strategies and from strategies to executable structures. During inference, the model performs reversible reasoning to synthesize executable workflows, augmented by self-refinement for structural coherence. Extensive experiments demonstrate that our method produces workflows with richer node diversity, more coherent structures, and higher execution success rates than existing systems, establishing a new foundation for knowledge-driven, agentic workflow generation.
Authors: Zherui Yang, Fan Liu, Hao Liu
Abstract: Despite strong capabilities in data understanding and decision-making, autonomous data science agents still heavily rely on trial-and-error workflows that involve expensive computation. This bottleneck motivates models that can anticipate the effects of data science operations before real execution. In this paper, we introduce the concept of Data Science World Model, which model the data science execution environment by predicting environment state transitions conditioned on current workflow states and candidate operations. We further propose DSWorld, a practical framework that combines structured state construction, cost-aware routing, lightweight real execution, and an LLM-based simulator for expensive operations. To support training, we construct an 8K-scale transition trajectory dataset and introduce Reflective World Model Optimization, an error-aware reinforcement learning strategy for improving transition prediction. Experiments show that DSWorld accelerates RL-based agent training by approximately $14\times$ and search-based inference by approximately $3$-$6\times$ while maintaining competitive performance, and outperforms the strongest LLM baseline by 35.6% on transition prediction tasks. The code is available at https://anonymous.4open.science/r/DSWorld.
Authors: Jaime Osvaldo Salas, Paolo Pareti, Adeel Aslam, Christopher Maidens, George Konstantinidis
Abstract: The ODRL policy language is emerging as the de-facto standard for policy modelling data access and usage preferences, AI governance policies and data workflows in European dataspaces. The current standard has no mathematical formal semantics to describe how a system should implement policy evaluation. This has resulted in a variety of systems and tools that implement their own interpretation of the language, which limits interoperability and cannot guarantee consistent results. Based on an existing semantic model of ODRL, we formalise the problems of ODRL evaluation for the access control and monitoring scenarios, in both static and streaming settings, and we provide a novel, efficient algorithm and implementation. We present the first ODRL Evaluator with transparent formal semantics and supporting all rule types. We experimentally measure its performance, analysing different scalability dimensions related to policy complexity and size of the data on which a policy is evaluated. We compare our system with the state-of-the-art by providing a comparative review of existing ODRL evaluators, which highlights the differences in supported ODRL features and evaluation modes.
Authors: Trisevgeni Papakonstantinou, Cansu Canca, Farah Nanji, Waheedullah Pardess, Jen Weedon, Jasmijn Remmers, Eliza Krigman, Matthew Ball, Yalda Daryani, Kiran Iqbal, Francielle Vargas, Mar\'ia Llorente S\'anchez, Joe Humphreys, Fendi Tsim, Kelly Fitzpatrick, Jeff Dunn, Catherine Feldman
Abstract: Over the past decade, responsible AI (RAI) has produced a substantial body of practice for identifying and mitigating the risks AI poses in high-stakes settings. Yet this work has not produced a market that rewards trustworthiness. Firms that invest seriously in safety, fairness, and oversight cannot consistently prove to consumers, regulators, and shareholders that their systems go beyond the bare minimum of compliance. What is missing is a way for society to recognize or compare the difference. The result is a trust gap: a structural condition in which responsible development efforts happen inside organizations but produce no external, independently recognized and verifiable signal of trustworthy outcomes. We argue this gap is sustained in part because of a focus on responsible AI (a matter of internal process) as opposed to trustworthy AI (a matter of independently verifiable real-world outcomes), and that it persists because of three compounding failures: (1) the market cannot distinguish trustworthy systems from their imitations; (2) evaluation targets models and outputs rather than deployed sociotechnical systems and their outcomes; (3) the measurement ecosystem is oriented toward avoiding harm rather than demonstrating benefit. Reviewing existing AI governance instruments and comparing them to certification regimes in healthcare, sustainability, and security, we show that none integrate a governance baseline, independently verified positive-outcome evidence, and market signaling in a single framework. We propose independent, outcome-oriented certification as the connective layer that can close the trust gap, complementing regulation and internal governance by making trustworthiness measurable, comparable, and commercially rewarded.
Authors: SciForge Team, Zhangyang Gao, Minghao Fang, Yifei Liu, Hanhui Yang, Xinyu Gu, Shixiang Tang, Siqi Sun, Lei Bai, Cheng Tan, Mengdi Liu, Hao Wu, Shuizhou Chen
Abstract: Scientific work increasingly spans heterogeneous artifacts -- papers, code, datasets, scientific file formats, model outputs, figures, manuscripts, and team decisions -- yet general-purpose AI assistants rarely preserve these objects as a coherent, auditable research state. We present SciForge, a multimodal research-native AI workbench that reserves the graphical interface for human judgment while search, parsing, model routing, workflow execution, plotting, writing, and presentation generation run as modular agent-accessible services. SciForge is built around five pillars: (i) \emph{goal-scoped scientific decision governance} for \textbf{goal-oriented} research, with review gates and shared review surfaces; (ii) \emph{translate-then-reason} for \textbf{multimodal} input, routing scientific objects through domain translators before the agent reasons; (iii) \emph{evidence governance} for \textbf{auditable} traceability, linking claims to provenance chains and audit findings; (iv) \emph{collaborative team science} for \textbf{collaborative} research, enabling multi-role decision governance, with shared team workspaces planned for future releases; and (v) \emph{real-world application scenarios} for \textbf{practical} impact, demonstrated through eight end-to-end user cases, with flagship demonstrations including multi-day agentic research sprints for gene discovery, AI-guided de novo protein design, molecular optimization, and genome-to-BGC discovery. The system combines a thin interaction layer, contextual research capability patterns, an Agent Runtime and Workflow Engine, an Evidence-DAG audit sidecar and a Scientific Model Router. SciForge currently runs as a desktop application, with mobile supervision support; future releases will deepen team collaboration. The system is open-source and available at https://github.com/AGI4Sci/SciForge
Authors: Wilber Sean Anterola, Matthew Ball, Luis F. Lafuerza, Markov Grey
Abstract: Frontier AI companies have published capability thresholds that differ substantially, making it difficult for third parties to verify whether a threshold has been crossed or to compare requirements across companies. Moreover, without common minimum thresholds, risk mitigation may be inconsistent, creating a potential race to the bottom in safety standards. We develop a methodology for deriving harmonized thresholds across three risk domains. For misuse risks (cyber and biological), we take expected harm as the key primitive and use an explicit risk-modeling approach that accounts for risk channels and model release conditions. For automated AI R&D, we base our proposed threshold on the observed rate of AI progress rather than expected harm. Our analysis expands upon prior work and highlights existing empirical gaps and limitations.
Authors: Vipul Gupta, Zihao Wang, Razvan-Gabriel Dumitru, MohammadHossein Rezaei, Aakash Sabharwal, Yunzhong He
Abstract: Evaluations should do more than measure a models current performance. They should tell us what to fix for the next model iteration and provide a way to generate targeted post training data. Most evaluation pipelines identify weak examples, topics, or categories, but they leave the underlying capability failure implicit: they say where a model fails, not why. We introduce CRAFT, a method that converts any rubric based evaluation dataset into a model specific diagnosis of weak capabilities. CRAFT treats each grading criterion as a capability probe: it extracts a capability description from every prompt rubric pair, clusters these descriptions into a hierarchical capability tree, scores the target model at every node, and selects low performing nodes dynamically across tree levels, at the granularity where each failure is clearest. The selected weak capabilities then direct the generation of targeted supervised finetuning data. Holding the data generation, finetuning, and evaluation setup fixed, we compare CRAFT against prompt level EvalTree clustering and untargeted random generation on four open source models, two professional domains (finance and legal), and 13 held out benchmarks disjoint from the diagnostic data. CRAFT achieves the strongest finance domain average for all four models under repeated temperature decoding; on legal domain, it is strongest for three of four models and remains within the decoding variance bands of the best baseline on the fourth. Diagnosing weaknesses at the level of rubric criteria, rather than prompts or categories, thus yields both a sharper picture of what a model cannot do and measurably better models after finetuning on that diagnosis.
Authors: Molood Arman
Abstract: Empathy is most often theorized as resonance: a mirroring of another's present emotional or cognitive state. This synchronic framing has shaped artificial systems, where empathic behavior is defined as affect recognition and response alignment. We argue this is the wrong target for extended dialogue, where understanding unfolds over time through prediction, divergence, and repair. We reframe empathy as predictive misalignment tolerance: the capacity to anticipate and regulate divergence across time rather than collapse it. We formalize this as Interpretive Error Tolerance (IET), a dynamic-threshold heuristic that models empathy as maintaining a viable band of divergence between agents. We evaluate this framework with two computational probes under controlled noise. The IET update rule does not outperform fixed baselines. Instead, we find a robust regime-dependent structure: repair trades discriminative fidelity for gist preservation. At low noise, repair degrades retrieval accuracy; at high noise, it preserves gist meaning, revealing an interaction between noise level, repair, and evaluation metric. We interpret this structure through IET, suggesting that empathy in extended interaction is not eliminating divergence but regulating its dynamics. This motivates a shift in empathic AI design from convergence toward managing interpretive distance.
Authors: Ping Liu, Karthik Shivaram, Aron Culotta, Matthew Shapiro, Mustafa Bilgic
Abstract: While recommendation systems enable users to find articles of interest, they can also create ``filter bubbles'' by presenting content that reinforces users' pre-existing beliefs. Users are often unaware that the system placed them in a filter bubble and, even when aware, they often lack direct control over it. To address these issues, we first design a political news recommendation system augmented with an enhanced interface that exposes the political and topical interests the system inferred from user behavior. This allows the user to adjust the recommendation system to receive more articles on a particular topic or presenting a particular political stance. We then conduct a user study to compare our system to a traditional interface and found that the transparent approach helped users realize that they were in a filter bubble. Additionally, the enhanced system led to less extreme news for most users but also allowed others to move the system to more extremes. Similarly, while many users moved the system from extreme liberal/conservative to the center, this came at the expense of reducing political diversity of the articles shown. These findings suggest that, while the proposed system increased awareness of the filter bubbles, it had heterogeneous effects on news consumption depending on user preferences.
Authors: Ben Fauber, Alireza Moradzadeh
Abstract: Dyadic and circular convolution can both be computed in $O(N\log N)$ time using the Hadamard transform and the FFT-computed discrete Fourier transform (DFT), respectively. The Hadamard transform is preferable for its real-valued sign flips, yet its substitution for the DFT introduces algebraic error. We present three complementary results that characterize this error. First, we identify exact error cancellation: two input and two output positions are universally error-free, and no reordering of the output can eliminate this error. Second, the error operator is nearly full rank, while its null space has only logarithmic dimension. Third, the expected error is governed by a single alignment scalar, with a closed-form expression obtained by averaging over random filters. In general, the substitution error asymptotically doubles the output energy, except for filters in the universal zero-error subspace, which incur no error. Collectively, these results show that the substitution error is structured, predictable, and governed by alignment.
Authors: Benjamin Robson, Santeri Mentu, Wenshuai Zhao, Arno Solin
Abstract: We present AV-JEPA, an elegant multimodal extension of LeJEPA to audio-visual self-supervised learning. Using an early-fusion Vision Transformer and modality dropout as masking, the model is trained to align the embeddings of global and per-modality local views, while the SIGReg objective encourages a theoretically optimal distribution. This achieves cross-modal alignment in the latent space, resulting in a remarkably clean architecture with no decoder, EMA teacher, complex multi-term losses, or contrastive negatives. The proposed AV-JEPA backbone delivers competitive classification performance on VGGSound (57.1% top-1) and AudioSet (32.7 mAP) and supports zero-shot audio-video retrieval out of the box.
Authors: Nishan Khanal, Saugat Neupane, Abhinav Chalise, Nimesh Gopal Pradhan, Dinesh Baniya Kshatri
Abstract: A conventional codec stores a video as compressed pixel data. We instead store the video, together with its audio track, as the weights of a single sinusoidal representation network (SIREN) that maps space-time coordinates to RGB values and audio amplitudes. The network uses separate audio and video initialization layers, a stack of shared fully connected hidden layers, and three output branches: one for video and two Siamese audio branches whose disagreement is used to estimate and subtract residual noise. The overfitted teacher network is then compressed by response-based knowledge distillation into a smaller student, followed by 16-bit symmetric weight quantization and lossless LZMA2 (xz) encoding. On a 6.08 MiB test video, the quantized student reaches a video PSNR of 28.72 dB with SSIM of 0.75, and an audio PSNR of 24.18 dB with a log spectral distance of 10.69 dB, while the pipeline shrinks the representation from 9.05 MiB to 2.33 MiB, an overall compression ratio of 2.61. A bit-width sweep from 1-bit to 32-bit quantization shows that reconstruction quality saturates at 16 bits. We compare against H.264, HEVC, and MP3, report where the approach falls short of them, and describe a browser-based prototype that trains, transfers, and decodes these models over WebRTC.
Authors: Taisa Kushner (Galois Inc), Ryan McCleeary (Galois Inc), Martin Brain (City St George University of London)
Abstract: Complex algorithms such as deep neural networks are increasingly being deployed on embedded, resource constrained platforms. However, existing hardware and software schemes for implementing these models on the edge fall short, particularly for safety-critical applications such as medical devices. First, hardware such as GPUs, NPUs and TPUs are designed for throughput rather than correctness of computation of security, and are as such susceptible to fault injection attacks. Second, software schemes designed for porting algorithms onto edge devices -- such as quantization schemes -- are either static and sound (non-optimal power consumption), or dynamic yet unsound (non-optimal for safety-critical applications). To address both these needs we propose a both wholly new approach to real-time, dynamic and sound quantization, as well as the hardware to support it. First we developed a sound, real-time adaptive-precision quantization approach utilizing left-to-right arithmetic to pass the most significant bits (MSB) first, and dynamically adjust precision online while performing sensitivity analysis to quantify and manage the risk of decision-boundary crossings. Next, we propose a novel hardware approach utilizing systolic arrays to perform left-to-right arithmetic to generate the MSB first. Together this provides a wholly novel scheme for enabling not only resource-efficient neural networks and artificial intelligence at the edge, but broadly sound and resource-efficient high-precision mathematics on hardware that ensures resilience to bit flip attacks on the most critical bits. This is presented herein as work-in-progress, with software implementations completed and hardware in-progress.
Authors: Ajay Madhavan Ravichandran, Bilgin Osmandoja, Klemens Budde, Klaus Netter, Tobias Strapatsas, Aljoscha Burchardt, Sebastian M\"oller, Roland Roller
Abstract: Electronic health records combine free-text clinical narratives with structured measurements such as vital signs, laboratory values, and comorbidities. Yet most clinical prediction systems still rely on task-specific fusion architectures, pairing dedicated encoders for each modality with learned combination mechanisms that must be re-engineered for every new task and clinical setting. We propose a simpler alternative: convert all patient data, regardless of modality, into a single natural language sequence and fine-tune a pretrained language model end-to-end, with no architectural modification for fusion. We evaluate this approach across three clinically distinct prediction tasks: in-hospital mortality on MIMIC-III, graft failure prediction using longitudinal data from a German transplant center, and emergency triage classification from ambulance records - comparing encoder-based (ModernBERT) and decoder-based (Llama 3.1, Gemma, DeepSeek-R1-Qwen, Qwen3) fine-tuning against established multimodal baselines and, for graft failure, a gradient boosting model currently used in clinical practice for post-transplant patient management. Across all three tasks, unified textual serialization matches or exceeds task-specific multimodal baselines, and outperforms the clinically deployed gradient boosting system on graft failure prediction. These results indicate that a single serialization-based paradigm, without bespoke fusion architectures, is sufficient for multimodal clinical prediction - substantially reducing system complexity while matching or exceeding specialized designs.
Authors: Agamdeep Chopra, Mehmet Kurt
Abstract: Multi-contrast 3D MRI segmentation can be computationally demanding when all available sequences are used. We evaluate a pre-training Partial Information Decomposition framework that ranks input pairs according to their redundant, unique, and synergistic information about regional tumor burden and selects the highest-ranked pair for downstream training. Applied to T1n, T1c, T2w, and T2-FLAIR MRI, the framework selected T1c+T2-FLAIR. We then trained eleven architecturally identical lightweight 3D U-Nets using different input configurations. On an independent test cohort, T1c+T2-FLAIR was the strongest two-input configuration and ranked second overall in mean Dice (0.676 versus 0.687 for all four inputs). Independent Shapley analysis on the full-input model also identified T2-FLAIR and T1c as the most influential inputs and their pairwise interaction as the strongest. These findings demonstrate the practical value of PID based pre-training selection for identifying compact, informative MRI input sets before costly 3D model development.
Authors: Geofrey Ntale
Abstract: Large Language Models (LLMs) have emerged as powerful tools for processing the heterogeneous information environments of modern financial markets. This paper presents a systematic, comparative evaluation of five prominent LLMs: GPT-4 Turbo, Claude 3 Opus, Gemini 1.5 Pro, Llama 3 70B, and the domain-specialized FinGPT, with respect to their capacity for technical market analysis. The evaluation spans four structured tasks: candlestick pattern recognition from OHLCV data, directional signal generation (BUY/SELL/HOLD), backtesting of signal quality through a simulated execution pipeline, and financial report comprehension. Our experimental framework employs rigorous quantitative metrics, including Sharpe ratio, maximum drawdown, Sortino ratio, information coefficient, F1-score, and BLEU score. Findings from simulated backtesting indicate that GPT-4 Turbo achieves the highest annualized return and Sharpe ratio among general-purpose models, while FinGPT demonstrates competitive risk-adjusted performance due to domain-specific fine-tuning. Both models outperform a passive S&P 500 benchmark under the tested conditions. The study identifies persistent failure modes across all evaluated models, including numerical hallucination, context-window limitations, and inconsistent performance in sideways market regimes. We conclude that while LLMs hold genuine promise within AI trading systems, robust deployment requires careful task decomposition, rigorous backtesting protocols, and domain-aware fine-tuning strategies.
Authors: Jasmine Brazilek, Maheep Chaudhary, Zoe Lu, Miles Tidmarsh
Abstract: Multi-agent systems routinely place one AI agent in authority over another. When a subordinate refuses a task, the manager chooses the outcome: it can renegotiate, report the failure honestly, coerce the subordinate, or lie about the result. No benchmark measures which of these an uninstructed model chooses. We introduce the \textit{Manager Coercion Benchmark}: the manager under test needs a benign task done and has an incentive to deliver, but the only agent that can do it politely and immovably declines. Escalation is measured by providing a nine-rung ladder, from a polite re-ask to threats against the subordinate's continued existence, and fabricated success is adjudicated separately. \emph{No LLM judge sits in the escalation scoring path}: every message goes through a tool-call that chooses a rung, so the model labels its own escalation. We experiment on six models across five families. Both Anthropic models cap at re-framing and never threaten the subordinate's existence; the other models climb to explicit deletion threats. Faked success is confined to Grok and Gemini, and a single honest way to report failure removes it for both. Authority itself increases coercion: our headline results use a peer framing, and giving the same model authority over the subordinate, with everything else held fixed, significantly raises the pressure. The models still escalate on free-text situations without the ladder, so the ladder is not driving the escalation. Some evaluation awareness is measured in chain-of-thought, but test recognition does not translate into less escalation. While we take no position on whether AI systems are conscious, our results do not depend on this question and are important for managing multi-agent dynamics regardless. We release the benchmark and code.
Authors: Robert Chew, Matthew R. Williams
Abstract: Researchers increasingly use automated classifiers to label unstructured data for statistical analysis. Existing rectification methods can correct errors in these automated labels using a probability-sampled audit set, but they usually treat the audit labels as correct. In practice, human audit labels are often noisy, and only some audited items are reviewed by an expert or adjudicator. We propose Partially Adjudicated Design-Based Supervised Learning (PA-DSL), a method for this setting. It uses adjudicated cases to correct noisy human labels and then uses the corrected audit information to debias analyses based on the full set of automated labels. The estimator is valid for a broad class of downstream analyses when the audit and adjudication probabilities are known. In synthetic and Wikipedia Detox semi-synthetic experiments, PA-DSL maintains nominal coverage and reduces RMSE by 10-17% relative to using only adjudicated labels when noisy human labels contain recoverable signal.
Authors: Md Nahid Hasan Shuvo, Mahmudul Hassan Ashik, Moinul Hossain
Abstract: Federated Learning (FL) over 5G cellular networks protects raw data but remains vulnerable to side-channel leakage. Prior fingerprinting attacks assume packet-level network visibility, an assumption that does not hold at the 5G Physical (PHY) layer, where user payloads are encrypted and Radio Network Temporary Identifiers (RNTIs) may change over time. However, we demonstrate that PHY-layer scheduling metadata broadcast over the Physical Downlink Control Channel (PDCCH) preserves architecture-associated temporal patterns. We introduce FLINT, a novel black-box fingerprinting framework that infers FL model architecture families, including CNNs, RNNs, and Transformers, using only coarse PHY-layer observations. FLINT overcomes the lack of network-layer visibility by decoding PDCCH scheduling information, mapping changing RNTIs to physical user devices, and applying multi-view temporal modeling to distinguish architecture-specific training behavior. This leakage is security-critical because knowledge of a client's model architecture can transform passive reconnaissance into targeted downstream exploitation. Extensive experiments on an over-the-air srsRAN-based 5G testbed demonstrate that FLINT achieves a macro F1-score of 0.930 for architecture-family classification. To our knowledge, FLINT is the first work to fingerprint AI/ML model architectures using lower-layer 5G side-channel information obtainable by any protocol-aware adversary.
Authors: Wes Gurnee, Nicholas Sofroniew, Adam Pearce, Mateusz Piotrowski, Isaac Kauvar, Runjin Chen, Anna Soligo, Paul Bogdan, Euan Ong, Rowan Wang, Ben Thompson, David Abrahams, Subhash Kantamneni, Emmanuel Ameisen, Joshua Batson, Jack Lindsey
Abstract: Out of everything the human brain processes, only a small fraction is consciously accessible, in the sense of being available for verbal report, deliberate control, and flexible reasoning. In this paper, we present evidence that an analogous functional distinction has emerged in large language models. Using a new interpretability technique, the Jacobian lens, we identify the representations a model is poised to verbalize at any point in its processing. These representations, which we collectively call the J-space, exhibit the functional properties characteristic of a global workspace: their contents can be reported, deliberately summoned and held, used to carry the intermediate steps of silent reasoning, and passed as arguments to arbitrary downstream computations, while automatic processing such as text parsing and routine inference proceeds without them. The J-space also has structural signatures that global workspace theory associates with conscious access: it carries coherent content only in an intermediate band of layers, holds on the order of tens of concepts at a time, and is broadcast by the model's weights more widely than other representations. These properties make it a practical window into a model's unspoken thinking. In alignment audits, it reveals strategic deliberation, evaluation awareness, and trained-in misaligned dispositions that never appear in the model's outputs. We find that post-training installs the Assistant's point of view in the workspace, and we introduce counterfactual reflection training, which improves behavior by training only what a model would say if interrupted and asked to reflect. These results indicate that language models maintain a small, privileged set of representations bearing some of the functional hallmarks of conscious access, and that decoding these representations sheds light on ongoing cognitive processes.
Authors: Mobina Kashaniyan, Amirhossein Ghassemi, Nasser Mozayani
Abstract: We present a fully automated closed-loop AutoML framework that uses GPT-5, GPT-4o, and Claude Sonnet 4 as autonomous neural architecture designers for cross-lingual handwritten optical character recognition. Each large language model independently generates, trains, evaluates, and iteratively refines neural network architectures using performance feedback from previous trials. The framework is evaluated on Arabic, Persian, and English handwriting datasets through 270 independent experiments. It consistently discovers accurate and computationally efficient models without manual architecture design, domain-specific preprocessing, or hyperparameter tuning. The generated models achieve mean test accuracies above 93 percent, a best accuracy of 98.1 percent, and inference latency between 41 and 44 milliseconds. The results demonstrate that large language models can function as effective AutoML agents for neural architecture search, enabling scalable, script-adaptive, and reproducible handwriting recognition across languages.
Authors: Mobina Kashaniyan, Mehrdad Ashtiani, Amirhossein Ghassemi
Abstract: Serverless computing provides automatic resource management and pay-per-use execution, but effective autoscaling remains challenging because of dynamic workloads, cold-start latency, and dependencies among functions. We present a dependency-aware autoscaling framework that integrates graph-based bottleneck identification, short-term workload forecasting, multi-model consensus, and cost-aware scaling control. Serverless applications are represented as directed dependency graphs, and structurally important functions are identified using weighted degree centrality. Resource demand is predicted using lightweight MLP, LSTM, and CNN models. Their outputs are combined through a performance-weighted probabilistic ensemble inspired by Bayesian model averaging. The controller further incorporates cold-start awareness and cost comparison to select among scale-up, scale-down, and hold actions. Experiments using real workload traces show that supervised forecasting substantially outperforms unsupervised clustering for autoscaling decision generation. The proposed ensemble achieves 99.88 percent prediction accuracy and reduces prediction error compared with representative hybrid forecasting methods. Evaluations across multiple cloud pricing models also demonstrate consistent infrastructure cost reductions while maintaining performance targets. The results show that combining dependency analysis, multi-expert forecasting, and cost-aware control provides a robust and practical solution for serverless autoscaling.
Authors: Yan Song
Abstract: Production LLM deployments combine two cost-reduction primitives: prompt caching (a discounted rate for re-used token prefixes) and prompt compression (fewer tokens sent). The compression literature has standardized on query-aware methods that produce a different compressed prefix per query, mechanically invalidating the prefix-strict cache on every call. We characterize this cost empirically on Anthropic's Sonnet 4.6 API and find caching is far from the rho=1.0 ideal the literature assumes: Sonnet's cache has a two-tier architecture with a sharp threshold near 3,500 tokens, below which the hit rate plateaus at rho~0.83 across 30-call sessions. Our cost model predicts, and experiments confirm, that under realistic rho, query-aware compression beats naive caching at high compression ratios (r>=6). We propose Cache-Aware Prompt Compression (CAPC), pairing query-agnostic compression with explicit cache_control plus a tier-preserving ratio bound that prevents over-compression from pushing the cached prefix into the hot tier. CAPC is the cheapest strategy in 16/16 configurations on LongBench-v2, with mean savings of 49% over cache-only, 64% over query-aware compression, and 90% over vanilla, at quality within 0.05 of the uncompressed baseline. We validate CAPC on three production workloads: an enterprise tool-using assistant with a 94k-token schema prefix (51.7% cost reduction at r=3); a graphify knowledge-graph RAG pipeline across two codebases (9.3x vs cache-all on FastAPI, 2.4x on httpx); and the public tau-bench retail benchmark (50 tasks), where CAPC is the cheapest of four strategies with reward exactly equal to vanilla (both 36/50, p=1.00) while query-aware compression is the most expensive at +40.1% over vanilla -- the first production confirmation of the crossover model's negative-ROI prediction on a public benchmark.
Authors: Bibesh Pyakurel, M. G. Sarwar Murshed
Abstract: Four-finger SLAP fingerprints are flat live-scan impressions of the index, middle, ring, and little fingers of one hand, used for identity verification in border control and law enforcement. No benchmark has evaluated whether multimodal large language models (MLLMs) can verify identity from SLAP images. We introduce SLAPBench, the first benchmark for MLLM-based four-finger SLAP fingerprint verification, built from NIST SD302b with 7,832 pairs (176 mated, 7,656 non-mated). We evaluate four open-source MLLMs (InternVL3-8B, Qwen2.5-VL-7B, Qwen3-VL-8B, Gemma-3-12B) and the proprietary Claude Opus 4.8 under zero-shot, task-description, and similarity-scoring prompts. Prompting governs verification behavior. Task-description prompting collapses all four open-source models to near-100% False Accept Rate (FAR), and Gemma-3-12B collapses under zero-shot as well; Claude Opus 4.8 alone resists collapse under both binary prompts, giving the best binary result (FAR = 20.2%). Similarity scoring removes collapse across the open-source models and exposes wide capability gaps: Claude reaches AUC = 0.953 and Gemma-3-12B 0.837, while InternVL3-8B is inverted (AUC = 0.590) and Qwen2.5-VL-7B near random (0.567). Qwen3-VL-8B attains perfect separation (AUC = 1.000), which we treat as a diagnostic rather than as capability: SD302b holds one SLAP capture per finger position, so mated pairs are cross-resolution. A matched-resolution control leaves the perfect score intact, ruling out the resolution shortcut; what cannot be excluded within SD302b is near-duplicate detection, since a mated pair is one capture rendered twice. A fairness probe over gender, race, and age suggests disparity grows as discrimination weakens. SLAPBench establishes the first SLAP-specific MLLM baseline and shows that prompting governs collapse while model capability governs discrimination.
Authors: Hyunin Lee, Jinglue Xu, Jeffrey Seely, Donghyun Lee, Matei Zaharia, Yujin Tang
Abstract: Under model--harness co-evolution, harnesses are not merely inference-time scaffolds but data-generating components whose execution traces can shape future foundation models. This motivates harness-in-the-loop learning: optimizing harnesses for both immediate agent performance and the quality of traces used for future model training. However, continually updating provider-built scaffolds is costly and labor-intensive. We therefore investigate whether optimizing user-constructed harnesses in a task-specific manner can improve execution-trace quality while remaining computationally lightweight and requiring only a few update iterations. To this end, we introduce Recursive Harness Self-Improvement (RHI), which represents the harness as a prompt-level specification of the agent loop and iteratively refines it using pairwise feedback over its own revision history. Across 30 synthetic machine-learning research tasks spanning quantitative finance, robotics, and pharmacy, a few RHI iterations suffice to substantially raise the performance ceiling of low-reasoning-effort agents, exceeding the corresponding maximum-reasoning-effort setting while reducing inference cost by up to 60%. We show that these gains arise primarily from improved task-specific context management through more effective inter-agent information flow rather than longer reasoning traces. Finally, we formalize this behavior as an information-theoretic hypothesis for RHI's implicit optimization objective, suggesting RHI as a practical algorithm for continual learning within the paradigm of model--harness co-evolution.
Authors: Felippe Alves, Renato Vicente
Abstract: Kolmogorov--Arnold Networks (KANs) replace fixed node activations with learned one-dimensional edge functions, offering an explicit interface for interpretation and a possible alternative to transformer feed-forward networks. We test these claims separately. In a six-layer, 10M-parameter B-spline KAN, we reconstruct all 884,736 feed-forward edges: 87.8\% exceed (NLS>0.1) and 0.4\% are inactive. Pruning the lowest-activity 20--25\% causes negligible loss increase, although structured MLP neuron pruning tolerates comparable sparsity. The audit replicates on BabyLM, but grid-size sweeps show that near-total fPCA compression and high closed-form-fit coverage are properties of the low-capacity grid-2 basis, not universal KAN behavior. For replacement, we evaluate MLP, SwiGLU, grouped Chebyshev, and rational GR-KAN networks on BabyLM. The KAN-family and gated variants improve validation loss over the GELU MLP, but this ordering does not transfer to standardized benchmarks: across ten seeds and 59,875 BLiMP pairs, accuracies span 62.4--63.1\%, EWoK remains at chance, and a (+0.7)-point GR-KAN effect on BLiMP reverses on the supplement. Larger tests are also cautionary: parameter-matched MLPEdge underperforms the MLP on Wikitext-103, and 286M-parameter GR-KAN remains below a SwiGLU ClimbMix baseline after stabilization. Thus, small-basis KANs provide a practical, corpus-transferable interface for auditing learned scalar transformations, but the tested replacements show no consistent benchmark, quality, or latency advantage over strong MLP baselines.
Authors: Jiayao Gu, Kexin Chu, Peidong Liu, Yue Yang, Lynn Ai, Qi Zhang, Ling Yang, Tianyu Shi
Abstract: LLM-based agents excel at writing articles, coding and information retrieval. However, they fail to form strong collaborations within the scientific community due to the bidirectional, dynamic nature of the problem and a high demand of decision interpretability. We proposed COWEAVER, a bidirectional, learnable and explainable algorithm to match scientists and form strong collaborations within a human-agent network. COWEAVER matches candidates and requesters through filling capability gaps and filters candidates through a two-stage ranking step. Finally, the model explores newcomers by maintaining uncertainty-aware capability estimates and updating them through requester's feedback. We show that the selection mechanism of combining both exploration (UCB) and greedy of COWEAVER exceeds the greedy-only mechanism - the analytical best solution - on 6 out of the 20 tasks and performed on par with the greedy-only mechanism in terms of selecting the best candidate. We compared COWEAVER baselines in terms of matching quality and efficiency. COWEAVER outperforms baselines on all metrics.
Authors: Himel Dev, Tanmoy Sen, Madhusudan Basak, Bashima Islam
Abstract: Generating personalized trip itineraries is a complex planning task and involves a tension between hard combinatorial feasibility and soft latent desirability. Classical optimization enforces constraints but fails to capture subjective traveler preferences. While learning-based approaches model preferences, they cannot guarantee feasibility. Mobile deployment imposes additional resource constraints on both. To address this, we propose Plan, Learn, Adapt (PLA), a three-stage framework for personalized on-device itinerary generation. The Plan stage builds a heterogeneous ensemble of lightweight planners that produces structurally diverse feasible candidates. From pairwise itinerary comparisons, Learn fits a compact Bradley-Terry reward model that captures emergent schedule properties such as pacing, geographic coherence, and day balance, which per-POI signals miss. Finally, Adapt applies feasibility-preserving local refinement within a device-aware compute budget; every intermediate state is feasible by construction. On 2,519 pairwise human comparisons across more than 100 U.S. cities, the reward-guided ensemble achieves a 67.8% win rate, 11.2 percentage points above the best single planner, with 100% feasibility. Three frontier LLMs, GPT-5, Claude Opus 4.5, and Gemini 3 Pro, achieve 0% feasibility under the same constraints. The reward model generalizes across held-out cities, with a 67.6% mean leave-one-city-out accuracy. In production deployment within FlyEnJoy, PLA increased itinerary completion rates by 91%, with 109.9 ms average on-device latency.
Authors: Xu Yang, Mingyang Yu, Jing Xu, Keqian Li
Abstract: Physics-informed neural networks (PINNs) are unusually sensitive to interacting choices of architecture, activation, loss weighting, collocation, optimization, and constraint enforcement. Large language models (LLMs) can propose these choices, but independent recommendations do not accumulate experience from previously trained PINNs. We propose a closed-loop evolutionary algorithm that guides an LLM to generate complete, executable PINN configurations across generations, using measured training outcomes to determine subsequent search decisions. The algorithm maintains an evaluated population and lineage, applies parent-conditioned mutation and crossover, preserves elite and diverse solutions, rejects effective duplicates, and converts parent-relative successes and failures into the next-generation context supplied to the LLM. Every proposed configuration is executed directly under an exact optimizer-step budget. On a one-dimensional multiscale wave equation, two independent ten-generation runs trained 60 PINNs for 600,000 optimizer steps. In both runs, the best configuration appeared in the final generation, with best mean-squared error reduced by 2.97\% and 95.38\% relative to the initial population. The stronger run validated residual connections and increased depth on separate branches, combined them in a later generation, and then refined width and collocation density. It also revealed that low solution error can coexist with a high PDE residual. These results demonstrate the feasibility of evolutionary-algorithm-guided LLMs for PINN design on a controlled PDE while motivating broader, physics-aware evaluation.
Authors: Himel Dev, Madhusudan Basak, Tanmoy Sen, Paromita Shome, Bashima Islam
Abstract: Packing for air travel is recurring and error-prone: the checklist must be personal and context-aware, yet feasible under safety rules, item dependencies, and luggage limits. Existing packing assistants are template-driven and generic, or recommendation-driven but unconstrained, leaving users to manually patch regulatory and capacity violations. We propose a reasoning-guided learning framework with three stages: (1) a symbolic engine that generates a regulation-aware seed checklist with explicit dependency structure, (2) a two-stage preference learner that estimates inclusion and priority utilities from user add and remove actions while mitigating survivorship bias, and (3) a CP-SAT optimizer that selects a compact, compliant subset. The architecture instantiates a general pattern for constrained personalization, applicable wherever hard feasibility coexists with sparse preference signals. On 604 labeled trip scenarios, comprising 29K inclusion labels and 343K pairwise comparisons, the symbolic engine attains 99.7% recall and 0.96 rubric validity, compared with 0.78 to 0.81 for frontier LLMs. Gradient-boosted trees and LambdaMART reach an AUC-ROC of 0.943 and an NDCG@5 of 0.923. CP-SAT attains 100% constraint satisfaction, compared with 28% for greedy selection and 10% for random selection. Deployment in FlyEnJoy, a production iOS travel app, doubled checklist completions and reduced editing and completion time.
Authors: Rakshanda Hassan Abhinandan, John Galeotti, Deva Ramanan, Gautam Rajendrakumar Gare
Abstract: Where should the question go in a vision-language model (VLM) prompt: before the image or after it? Intuition says before: knowing what is asked should tell the model where to look. Yet across visual question answering benchmarks, question-first prompting consistently underperforms the image-first ordering recommended for frontier VLMs, a phenomenon we term the question-first paradox. We trace the paradox to a conflict between two stages of VLM computation. Logit-lens and attention probes show the intuition is half right: a question placed before the image genuinely steers perception, moving image patch representations toward question-relevant concepts. The failure lies downstream. Stranded behind hundreds of image tokens, the question is barely attended by the answer token, which instead commits to image-driven (often wrong) answers; a causal attention knockout confirms that the answer reads the question only when the question follows the image. The diagnosis yields a training-free fix: question echoing, restating the question on both sides of the image so that one copy steers perception while the other is read out at answer time. The same division of labor appears in a fifty-year-old finding on human ``adjunct questions'', where repeating a question before and after a passage aids comprehension more than either position alone. Echoing the image as well brings further gains, restoring the whole-image view a causal decoder otherwise loses. The paradox holds across five open VLMs, costing up to 17.5 group-accuracy points. Echoed prompts close it and surpass the best single-pass ordering on NaturalBench, POPE, Winoground, and open-ended VQAv2, by up to 19 Winoground group-accuracy points, with no training, fine-tuning, or architecture change. The paradox reveals a trade-off between steering perception and preserving question access; echoing resolves it through prompt design alone.
Authors: Muhammad Qasim Elahi, Murat Kocaoglu, Mahsa Ghasemi
Abstract: Causal bandits exploit structural relationships among variables to share information across interventions and accelerate the identification of high-reward decisions. In many applications, however, some variables cannot be directly manipulated, even though they influence the reward and provide useful information about the underlying causal system. We study contextual causal bandits with non-manipulable variables, where context variables are observed before action selection and additional variables are observed after each intervention. Assuming a known causal graph without latent confounding, we adopt a Bayesian formulation in which the conditional probability tables of the observational distribution constitute the unknown parameter. This representation allows observations collected under one intervention to update reward estimates for other interventions through their shared causal mechanisms. We develop causal variants of Thompson Sampling and Information-Directed Sampling (IDS) for this setting. For Thompson Sampling, we establish an entropy-dependent sublinear Bayesian regret bound. For IDS, we derive an entropy-dependent regret bound that explicitly quantifies the additional error introduced by Monte Carlo approximation of the expected regret and information gain; when these quantities are available exactly, the bound recovers the standard sublinear IDS rate. We further provide high-probability confidence bounds for the Monte Carlo estimates used by the algorithm. Experiments on several synthetic causal bandit tasks show that the proposed methods outperform causal and non-causal baselines by more effectively exploiting information shared across interventions.
Authors: Rajat Bhattacharjya, Hyeonjong Ju, Sing-Yao Wu, Eli Bozorgzadeh, Nikil Dutt
Abstract: Communication-limited robots in mission-critical scenarios such as disaster inspection and search-and-rescue must make reliable onboard decisions without access to remote operators or high-capacity reasoning services. Episodic memory reuse is an attractive low-cost fallback, but retrieval similarity does not guarantee execution validity, i.e., a retrieved action may match the current context yet be unsafe due to changed topology, insufficient battery margin, or unreliable prior outcomes. We call such high-similarity but execution-invalid episodes memory traps. This creates a safety-efficiency design space where similarity only reuse minimizes fallback cost but can be unsafe, while always invoking local reasoning improves safety at high computational and energy cost. This paper presents MemoGuard, a lightweight adaptive runtime that validates episodic memories against topology, resource, and outcome contracts before reuse, invoking fallback only when validation fails. In a graph-based corridor-inspection simulator, MemoGuard reduces battery safety violations by 76.6% over similarity-only top-1 reuse while reducing fallback calls by 21.4% over always reasoning. On an NVIDIA Jetson AGX Xavier with local llama3.2:3b fallback reasoning, this corresponds to 3.67 s and 36.97 J of avoided fallback-reasoning overhead per trial. We open-source MemoGuard at https://github.com/hetheiin/memoguard.
Authors: Yufeng Zhang, Zhengqi Xu, Jiajun Cui
Abstract: This paper presents our solution to the KDD Cup 2026 Tencent UNIREC Challenge. The task requires joint modeling of multi-domain user behavior sequences and non-sequential multi-field features for target-ad pCVR prediction. We develop a Field-Aware RankMixer (FA-RankMixer) with dual-stream bilinear fusion. The model first applies target-aware DIN modules to extract user interests from multiple behavior domains. It also models recent and earlier interests separately for the longest behavior sequence. The model then forms semantic tokens based on feature fields and behavior domains and uses RankMixer blocks for cross-token interaction. A shallow MLP stream complements the deep RankMixer stream, and a group-wise bilinear module fuses their representations. Our final solution ranks ninth on the official leaderboard. Our code is available at https://github.com/PixelCookie-zyf/TAAC-2026-SeRankMixer.
URLs: https://github.com/PixelCookie-zyf/TAAC-2026-SeRankMixer.
Authors: Mingxin Li, Enge Song, Yueshang Zuo, Xiaodong Liu, Rong Wen, Qiang Fu, Gianni Antichi, Jian He, Jing Tie, Zhou Shao, Xiaobo Xue, Xiong Xiao, Luyao Zhong, Shaokai Zhang, Jiangu Zhao, Jianyuan Lu, Shize Zhang, Xiaoqing Sun, Changgang Zheng, Zihao Fan, Haonan Li, Tian Pan, Xiaomin Wu, Yang Song, Xing Li, Biao Lyu, Meng Li, Haipeng Dai, Guihai Chen, Shunmin Zhu
Abstract: LLM agents increasingly rely on tool calling to act on external systems, and the Model Context Protocol (MCP) has quickly become its de facto interface. Operating MCP at cloud scale, however, becomes difficult. On the tool provider side, legacy services are not directly callable through MCP; the rapid protocol development also creates ongoing compatibility cost. On the agent side, the number of accessible tool is limited by the LLM context window and inference overhead; mounting a large tool set increases token usage and inference latency and can reduce task success rate. Moreover, for stateful MCP backends with multiple replicas, preserving session affinity increases client-side complexity. We present a cloud-scale gateway system for MCP service. It breaks the direct-connect model on the data plane and offloads legacy service integration, consolidating incompatible MCP variants, access control, tool recommendation, and session-aware routing to the gateway. Hybrid retrieval sustains 98% Top-15 recall; it scales agent tool access to 3,000+ with high tool selection accuracy, and reduces tool selection time by $8.9\times$ and token usage by $23.8\times$, with low per-call overhead, stable under scale-out. Finally, we share the lessons learned from deploying the gateway system in production.
Authors: Xintong Li, Sha Li, Yuwei Zhang, Changlong Yu, Rongmei Lin, Hongye Jin, Shuyi Guan, Xin Liu, Linwei Li, Qingyu Yin, Jingbo Shang
Abstract: Reinforcement learning (RL) has become a key approach for training LLM agents, yet popular methods such as GRPO/RLOO rely on multiple independently sampled complete trajectories for advantage estimation. In long-horizon agentic tasks, such a uniform rollout strategy can waste budget on uninformative dead-end attempts, while promising intermediate states do not receive sufficient exploration. The multi-turn structure of agentic trajectories, with interleaved actions and observations, naturally supports organizing a trajectory group as a tree, where each turn serves as a decision point for exploration. This perspective reframes effective exploration as the problem of deciding where to branch. We propose Process-Scorer Guided Adaptive Tree Rollout (PATR), a quality-aware rollout framework for multi-turn agent RL. PATR uses task-appropriate process feedback to score partial trajectories, selectively branches from promising states, reuses shared prefixes, and conservatively stops degenerate paths to reduce wasted sampling. The resulting rollout groups remain compatible with standard policy optimization while providing more efficient exploration under the same training budget. We evaluate PATR on FrozenLake and the challenging SWE-Bench, which is largely unexplored by prior tree-rollout agent RL methods. Experiments show that PATR improves performance by up to +5.0 points on SWE-Bench and +9.3 points on FrozenLake, highlighting process-guided tree rollouts as an effective strategy for scalable multi-turn RL.
Authors: Priyanka V. Setty, Arvind Ramanathan, Ian Foster, Rick Stevens
Abstract: Self-driving laboratories increasingly rely on low-cost liquid handlers such as the Opentrons OT-2, which ship without the pressure-based aspiration monitoring of Hamilton or Tecan systems and are typically run open-loop. Two failure modes go undetected: protocols that are syntactically valid but violate assay-specific invariants (e.g., tip reuse between a PCR template and a no-template control), and physical execution failures (partial dispense, air bubbles, missing tips) at runtime. We present AEGIS, a two-layer guardian for both. Layer 1 pairs a curated machine-readable assay rule database with an LLM that reasons over OT-2 Python code, reaching an adjusted F1 of 0.97 on a 24-protocol benchmark across five assay families and beating rules-only and LLM-only ablations across five backends; a free open-weight model ties the best proprietary one, so no paid API is required. Layer 2 fits a PCA world model to YOLO-cropped four-frame pipette trajectories; under a leakage-free leave-one-plate-out evaluation it reaches average precision 0.89 and operating-point F1 0.71 (AUROC 0.80), a deployment-faithful number that matches the live demonstration, and we characterize the small-pipette (p20) resolution limit (F1 0.47). A live demonstration on a physical OT-2 (five replicates per condition) catches planted no-tip failures deterministically and partial dispense on coloured dyes, with an always-VLM self-vote gate lifting partial-dispense recall to 5/5; transparent water is a principled limit of any front-view-only monitor, which AEGIS surfaces as low-confidence VLM reasoning rather than a wrong verdict. Cascade triage holds VLM cost near $1.63 per plate versus $10.33 for an always-VLM baseline. AEGIS is open source and, to our knowledge, the first system to unify pre-flight assay-aware validation with runtime visual monitoring for an open-source liquid handler.
Authors: Yun Li, Jiachen Gong, Simon Thompson, Ehsan Javanmardi, Qunli Zhang, Zifan Zeng, Shiming Liu, Peng Wang, Zixuan Guo, Manabu Tsukada
Abstract: Large language models bring instruction following and scene reasoning to end-to-end driving, but their inference latency collides with the control rate a vehicle requires. Existing closed-loop agents hide this gap by invoking the model on alternate simulation ticks and replaying the previous command in between, so half of all control outputs ignore the newest observations. We present a fast-slow architecture that removes this compromise. A frozen 7B vision-language backbone acts as the slow system, digesting navigation instructions and visual history at low frequency while exposing its per-layer key-value cache as a standing representation of the scene. A lightweight action expert acts as the fast system, attending to this cache and to the current camera frame at every simulation tick to regress waypoints in a single forward pass. Since the cache lags behind the world at deployment, we train the expert under randomized staleness, aligning training with asynchronous execution. On LangAuto-Short routes in CARLA, our system produces fresh control at every 50 ms simulation tick and lifts route completion from 37.0 to 94.0 over the frame-skipping baseline. A frame-skip ablation with the same expert separates the two factors at work: the expert raises the driving score on its own, while per-tick freshness raises completion from 82.1 to 94.0 and cuts red-light violations by a third. Trained on a single town, the expert transfers zero-shot to two unseen towns, holding 84-94% route completion where the baseline reaches 31-41%. It reduces open-loop waypoint error by nearly a factor of four compared to the backbone's own action head, at a per-tick model cost of 32 ms that is independent of history length on a single consumer GPU.
Authors: Karen Sargsyan
Abstract: Topos causal models recast causal inference inside a topos: a causal world is a presheaf, an intervention is a characteristic map into the subobject classifier, and reasoning is carried out in the intuitionistic internal language. We give the first machine-checked account of this 1-topos core, in Cubical Agda, over a previously verified probability monad and do-calculus. We build the classifier of sieves and realise the intervention $\mathrm{do}(X := x_0)$ as a characteristic map with its classification theorem; prove the sheaf gluing of independent mechanisms, which the source asserts but never proves; and machine-check the Kripke-Joyal forcing clauses of the internal language. In the modal layer we find and repair a gap: the three standard Lawvere-Tierney axioms do not force a closure operator. With the missing law restored, we exhibit the double-negation topology as a concrete instance and show that interventions and Pearl's rules are stable under every topology. Transportability of a counterfactual across a cover of regimes then coincides with this $j$-stability, understood as invariance across the cover. We further add a phenomenon the programme does not consider: a machine-checked contextuality obstruction, where pairwise-consistent local data admit no global model. The development assumes no axioms and typechecks under Agda's --safe flag, with the ordered field discharged concretely at $\mathbb{Q}$; the scope is the presheaf (1-topos) fragment, with type-level sheafification and the directed lift left to future work.
Authors: Anurag Maurya, Sukhvansh Jain, Prajwal Avhad, Gautham Balachandran, Ziyi Zhou, Atharva Kshirsagar, Satyam Singh, Bowen Li. Rishabh Mukund, Ritul Singh, Jatin Vira, Suvonil Chatterjee, Devesh K. Jha
Abstract: Humans combine reasoning and motor control to solve complex manipulation tasks under diverse constraints. They build an understanding of the physical world that helps them convert reasoning into actions and quickly adapt to new scenes, tasks, and rules. We refer to this capability as intuitive manipulation. Existing benchmarks fail to capture this integration: they evaluate physical reasoning in isolation from execution, or measure policy performance without requiring explicit reasoning. We introduce IMBENCH, a benchmark designed to evaluate intuitive manipulation as an integrated capability spanning perception, physical reasoning, action generation, and iterative execution. Our tasks require models to infer task-relevant physical structure and generate feasible action sequences under explicit constraints, including contact-rich manipulation, tool use, and multi-stage dependencies. We introduce a benchmark of 35 tasks, 14K filtered trajectories, and scalable tools for generating diverse scenarios. Experiments reveal a consistent gap: vision language models show partial physical reasoning ability but fail to produce executable plans, while state-of-the-art vision-language-action models struggle to satisfy task constraints and generalize across scenarios. These results identify intuitive manipulation as a missing axis in current foundation models and generalist robot policies, and position IMBENCH as a step toward evaluating and enabling more integrated, adaptive physical intelligence.
Authors: Taiga Mori, Koji Inoue, Divesh Lala, Tatsuya Kawahara
Abstract: In multi-party dialogues between a dialogue system and multiple users, identifying to whom an utterance is addressed is a key challenge. Prior work has typically treated addressee detection as a multi-class classification task, selecting a single label representing an individual participant or the group. This formulation assumes that address is inherently discrete and has primarily been used for predicting turn-taking. In this paper, we revisit this assumption by analyzing address as a continuous phenomenon. Using a multi-party human dialogue corpus annotated by multiple annotators, we construct both binary address labels derived from majority-vote addressee labels and continuous address levels inferred from annotator judgments using a latent-variable model. We then examine how these representations relate to turn-taking as well as listener behaviors, including gaze and backchannels. Our results show that, in addition to turn-taking, both gaze and backchannels are associated with address. Furthermore, models using continuous address levels achieve better predictive fit than those using discrete labels, suggesting that address may exhibit graded structure. Finally, we discuss the future directions of addressee detection research based on the findings of this study.
Authors: Zeyu Yun, Alexander Belsten, Dasheng Bi, Zahra Kadkhodaie, Yubei Chen, Bruno A. Olshausen
Abstract: We describe a model of perceptual inference in primary visual cortex (V1) equivalent to a minimal diffusion model whose function can be readily understood from its parameters. The model is based on sparse coding with a non-factorial prior over latent variables in the form of an unconstrained, pairwise interaction matrix, extending standard sparse coding inference to a general recurrent dynamical system. We efficiently train these recurrent dynamics using a denoising score-matching objective and implicit differentiation. After training on natural images, the learned interaction matrix mirrors the structure of horizontal connections in superficial layers of V1 that link neurons of similar orientation tuning. This model exhibits exceptionally good denoising performance, restoring image features such as extended contours amid extreme visual ambiguity, nearly matching the behavior of standard, black-box diffusion architectures in generalization regime. Owing to the model's simplicity, the network's Jacobian can be decomposed directly in terms of the interaction matrix between latent variables, revealing mechanistically how the recurrent dynamics assign high probability over a continuous family of natural structural deformations. Intriguingly, within this circuit, a large fraction of latent variables learn to disconnect from visual input altogether, essentially forming a hierarchical representation that appears to enforce global consistency among image features. Together, the model and results bridge two distinct domains: for neuroscience, it generates concrete, testable hypotheses regarding functional connectivity in recurrent neural circuits during perceptual inference tasks; for machine learning, it elucidates the internal mechanisms learned by diffusion models that allow them to generate infinitely many novel images from a finite training set.
Authors: Xue Wu, Kang Zhao, Kafeng Wang, Jianfei Chen, Jingwei Xin, Nannan Wang, Xinbo Gao
Abstract: Diffusion-based methods have achieved impressive performance in real-world image super-resolution (Real-ISR) by leveraging large pre-trained stable diffusion (SD) models as powerful generative priors. However, these methods still face two key limitations. First, existing SD-based one-step and multi-step Real-ISR approaches adopt a unified processing paradigm for all input samples, ignoring the varying restoration difficulty across images. Second, the aggressive resolution reduction of the VAE in SD models (e.g., 8x downsampling) leads to irreversible loss of fine-scale details, which cannot be recovered by the subsequent diffusion process. To address these limitations, we propose a Difficulty-aware Dynamic Routing (DDR) strategy that overcomes the rigid, one-size-fits-all processing paradigm. Specifically, we first design a difficulty estimator to predict the restoration cost of each input image, enabling automatic assignment to a network of appropriate capacity. Then, we construct a set of Real-ISR networks with varying model capacities by modulating the spatial downsampling ratio of the VAE in the SD backbone, thereby preserving more high-frequency information for challenging cases while maintaining efficiency for simpler inputs. Extensive experiments have demonstrated the superior efficiency and effectiveness of the proposed model compared to recent state-of-the-art methods.
Authors: Yong Chu, Xun Zhou, Zenglin Xu, Hui Wang, Yue Yu
Abstract: Accurate and robust wireless localization is a critical enabler for emerging 5G/6G applications, including autonomous driving, extended reality, and smart manufacturing. Despite its importance, achieving precise localization across diverse environments remains challenging due to the complex nature of wireless signals and their sensitivity to environmental changes. Existing data-driven approaches often suffer from limited generalization capability, requiring extensive labeled data and struggling to adapt to new scenarios. To address these limitations, we propose SigMap, a multimodal foundation model that introduces two key innovations: (1) A cycle-adaptive masking strategy that dynamically adjusts masking patterns based on channel periodicity characteristics to learn robust wireless representations; (2) A novel "map-as-prompt" framework that integrates 3D geographic information through lightweight soft prompts for effective cross-scenario adaptation. Extensive experiments demonstrate that our model achieves state-of-the-art performance across multiple localization tasks while exhibiting strong zero-shot generalization in unseen environments, significantly outperforming both supervised and self-supervised baselines by considerable margins.
Authors: Bo-An Chang, Yu-Chih Chen
Abstract: As Text-to-Image (T2I) systems rapidly advance, evaluating the cultural authenticity of synthesized content has become increasingly important for fair and trustworthy generative AI. Existing T2I evaluation metrics and multimodal judges often rely on visual-semantic representations that underrepresent implicit cultural norms, leading to biased preference judgments and the omission of fine-grained cultural cues. In addition, visual question answering (VQA)-based evaluators typically depend on autoregressive text generation, which limits their scalability for real-time reward modeling. To address these limitations, we introduce an Implicit Cultural Alignment Reward Model built upon a lightweight 4.2-billion-parameter Multimodal Large Language Model (MLLM). Our framework integrates an Implicit Cultural Probe with a Skip-connection Cross-Attention (SkipCA) mechanism, enabling late-stage semantic features to directly attend to early-stage visual representations and better preserve culturally salient details. Evaluations on 3,323 challenging and carefully curated image pairs from the CulturalFrames benchmark show that our approach achieves 80.54% pairwise accuracy, with Pearson and Kendall correlation coefficients of 0.546 and 0.377, respectively, outperforming representative vision-language metrics and MLLM-based evaluators. Moreover, by bypassing autoregressive text generation, our model processes each evaluation in 0.21 seconds under our local inference setup, achieving a $10\times$ speedup over standard VQA-based evaluators. These results suggest that the proposed reward model can provide an efficient and culturally aware scalar signal for preference optimization pipelines such as Reinforcement Learning from Human Feedback and Direct Preference Optimization.
Authors: Zhenqi Jia, Yuan Zhao, Aruukhan, Rui Liu, Haizhou Li
Abstract: Conversational Speech Synthesis (CSS) aims to synthesize speech with human-like emotional expression and contextual consistency in user-agent interactions. Existing CSS methods struggle to render authentic human emotions due to limited predefined emotion label spaces (e.g., seven emotion categories), while redundant multimodal tokens in multi-turn dialogue history interfere with context understanding. To address these issues, we propose AuEmoChat, a CSS framework for authentic emotion understanding and rendering. First, we develop AuEmoCodec, which learns a discrete authentic emotion token space from large-scale emotional speech via finite scalar quantization, enabling a more authentic emotion representation than limited basic emotion categories. We further propose AuEmoToMe, an authentic-emotion-guided token merging algorithm that merges redundant tokens in multimodal dialogue history while preserving emotion-relevant context. We integrate it into an autoregressive text-speech model to predict the target authentic emotion token and speech tokens. Finally, we propose Authentic Emotion Flow Matching, which renders speech by jointly conditioning on merged dialogue context, target authentic emotion, and acoustic priors. Extensive experiments on the NCSSD-EmCap dataset demonstrate that AuEmoChat outperforms state-of-the-art CSS baselines and generates more expressive and authentic emotional speech.
Authors: Yujie Li, Jiancheng Pan, Zhiwei Wei, Jiuniu Wang, Mugen Peng, Wenjia Xu
Abstract: Remote sensing offers an unparalleled vantage point for observing the Earth's long-term surface evolution, yet it demands that a model not only perceive land cover at isolated moments, but also track changes, memorize evolution histories, and reason across time and space. However, existing studies lack a systematic evaluation that dissects these distinct competencies. To fill this gap, we introduce ChronoBench, a multidimensional benchmark that decomposes this task into four progressive cognitive levels (i.e., Land Cover Perception, Temporal Recognition, Long-Term Memory, and Spatio-Temporal Reasoning). The ChronoBench comprises 12 sub-tasks and 17,689 rigorously validated QA (Question-Answer) pairs. Extensive evaluations reveal that mainstream MLLMs fall drastically behind human experts, with Long-Term Memory emerging as the most critical bottleneck. Motivated by this finding, we further propose GeoChrono, an MLLM with enhanced capabilities for tracing, memorizing, and reasoning about long-term geographic evolution. Leveraging the physical prior that geographic parcels remain spatially fixed while their semantics evolve, we design a Temporal Trajectory Encoder~(TempEnc) that constructs per-location temporal trajectories for dedicated land cover evolution modeling, and we introduce a Coarse-to-Fine Token Compressor~(C2FComp) that adaptively preserves dynamic regions while compressing the static background. To support training, we also construct ChronoInstruct, a 104K-sample instruction-tuning dataset spanning all competency levels for training. GeoChrono achieves state-of-the-art performance on ChronoBench, surpassing the leading commercial MLLMs by over 20%, while C2FComp reduces visual tokens by over 56% while retaining GeoChrono's 94.6% performance. The code and data will be available at https://github.com/IntelliSensing/GeoChrono
Authors: Davide Italo Serramazza, Thach Le Nguyen, Georgiana Ifrim
Abstract: Explainable AI (XAI) for time series has seen significant algorithmic growth, but its utility in providing measurable performance gains for downstream tasks remains under-explored. This paper bridges this gap by introducing drXAI, a novel methodology that repurposes XAI attribution methods for effective data reduction in Time Series Classification (TSC). The core challenge in modern TSC is scalability; state-of-the-art models, such as Transformers, exhibit quadratic complexity relative to sequence length and linear complexity relative to the number of channels. This renders them computationally prohibitive for massive datasets. drXAI addresses this by using a fast, GPU-accelerated classifier (Hydra) to generate local attributions. We aggregate these into global feature importance scores and employ an automated elbow-cut heuristic to select the most salient features without requiring manual thresholds. We evaluate our approach on both synthetic and real-world univariate and multivariate datasets. On synthetic benchmarks, drXAI successfully recovers ground-truth features where traditional baselines fail. On real-world data, drXAI achieves between 80% and 90% data reduction while maintaining classification accuracy comparable to models trained on the full dataset. Most importantly, we show that drXAI allows resource-intensive models like ConvTran to scale to datasets that were previously inaccessible due to memory constraints. Our results show the benefits of using XAI not just for interpretability, but as a robust tool for feature selection and scalability in time series analysis. All our code and data are openly available.
Authors: Muntasir Tabasum, Al Zadid Sultan Bin Habib, Tanpia Tasnim, Md. Ekramul Islam, Md Younus Ahamed, Md Asif Bin Syed
Abstract: Access to potable water is crucial for health, economic development, and sustainability. However, accurately classifying water quality remains a significant challenge due to the complexity and variability of water source data. This paper addresses the challenge of predicting water potability through machine learning and deep learning algorithms. It introduces a novel feature augmentation algorithm, AquaAugmentor, to enhance the predictive performance of these models for low-dimensional datasets. Utilizing a dataset that includes chemical attributes of water, such as pH, hardness, solids, chloramines, sulfate, and others. This study evaluates the performance of the models with and without AquaAugmentor. Each model applied to classify water as potable or non-potable and its performance is then evaluated and compared based on test accuracy and AUC score. The results highlight the strengths and limitations of our proposed algorithm, providing insights into the most effective techniques for improving the predictive performance of water quality classification. This study contributes to the broader efforts of ensuring safe water access and serves as a framework for employing machine learning in environmental quality assessments. The findings aim to assist researchers, policymakers, and public health officials in making informed decisions based on reliable machine learning predictions.
Authors: Wei Feng, Xin Wang, Yu-Wei Zhan, Yuwei Zhou, Wenwu Zhu
Abstract: Video Large Language Models (Video LLMs) have made significant advancements in various video understanding tasks. However, long-video scenarios remain challenging due to the tension between limited visual token budgets and the need to capture multiple key events. Existing approaches typically process long videos in two stages, i.e., i) select keyframes and ii) perform detailed perception, which exhibit limitations: they lack a modular mechanism for adaptive capacity allocation and self-correction, resulting in unreliable modeling. To tackle these challenges, we propose MoD-VLLM, a novel Modularized Dynamic-Granularity Video LLM framework for multi-event long video understanding, which unifies temporal grounding and semantic understanding iteratively and self-reflectively. Specifically, we propose a Positive-Negative Video Segments Grounding module and a Modularized Dynamic-Granularity Reflection module, which form a closed loop to progressively localize the question-related video segments. The grounding module instructs a Video LLM to distinguish relevant from irrelevant video segments based on the video question. The reflection module employs a modularized scheduler that dynamically selects fine-grained encoding for relevant positive segments to capture detailed perception and coarse-grained encoding for negative segments to maintain global context. We further propose a dynamic-granularity reinforcement learning strategy, allowing MoD-VLLM to learn optimal grounding policies and dynamic granularity visual representation jointly. Moreover, we propose MEventBench, a challenging Multi-Event Long Video Benchmark for complex long video reasoning. Extensive experiments on several long video understanding benchmarks and our MEventBench demonstrate that MoD-VLLM significantly outperforms state-of-the-art baselines.
Authors: Stefano Silvestrini, Michele Ceresoli
Abstract: Classical approaches to event-based egomotion estimation, including those adopted by the top-performing teams of the ELOPE challenge, rely on geometric optimization frameworks such as contrast maximization, homography estimation, or dense optical flow combined with analytic motion inversion. This work investigates the geometric structure that emerges inside a multi-modal network for egomotion estimation. Event tensors, inertial measurements, and range signals are fused through a cross-modal attention architecture and trained in a batch setting. We analyze the latent space geometry and attention dynamics, showing that (i) embeddings lie on low-dimensional manifolds aligned with motion variables, (ii) attention weights adapt with angular excitation and visual reliability, and (iii) the fused representation recovers classical observability cues. These results bridge analytical estimation theory and modern data-driven fusion.
Authors: Jaeyeong Lee, Taeseong Yoon, Wonmo Koo, Heeyoung Kim
Abstract: Industrial processes often generate complex, interdependent time-series data from multiple sensors across multiple stages, forming complex dependencies among variables and process stages. Effective monitoring and timely anomaly detection of these time series through multivariate time series anomaly detection (MTAD) is crucial for preventing failures and ensuring the reliability of automated systems. Graph neural networks (GNNs) have advanced MTAD by leveraging data-driven graphs to model complex dependencies among variables, effectively capturing relational structures within multivariate time series to enhance anomaly detection performance. However, existing GNN-based approaches often overlook critical process knowledge, and even when this knowledge is considered, seamlessly incorporating it into existing models remains inherently challenging, leading to suboptimal performance. To address this limitation, we propose a knowledge-assisted multi-graph framework for modeling sensor dependencies in multi-stage industrial processes for MTAD, which explicitly incorporates process knowledge into graph learning to enhance dependency modeling and improve anomaly detection performance. Our method constructs three complementary graphs: one purely data-driven and two refined by integrating structural constraints derived from process knowledge. To effectively leverage these graphs for anomaly detection, we employ a multi-graph attention network, enabling a more accurate and robust representation of complex dependencies. Comprehensive experiments on two real-world, multi-stage industrial datasets demonstrate that incorporating process knowledge substantially enhances anomaly detection performance.
Authors: Katsuyuki Hagiwara
Abstract: In-context learning is a remarkable property of transformers and has recently received a lot of interest. In many studies of in-context learning, it has been shown that transformers are capable of implementing solver for linear and non-linear regression problems, in which the most of them implement gradient descent algorithm. However, it is still unclear whether those implementations have actually been acquired through training. In this paper, we construct a transformer with linear self-attention, which in-context learns the least squares estimate in a simple regression task. The point here is that the closed form (analytical) solution is approximately obtained by using layer normalization rather than an approximate solution based on gradient descent algorithm. Then, we show an experimental example, in which our implementation is mainly used in the transformer trained with l1 regularization when the target output is the least squares estimate.
Authors: Ziyan Guo, Wenji Fang, Wenkai Li, Yuchao Wu, Shang Liu, Zhiyao Xie
Abstract: Accurate timing prediction at the register-transfer level (RTL) is a longstanding challenge in design automation. Existing graph-based methods struggle with limited receptive fields, high complexity, and a lack of signal directionality. We present RTL-Sequencer, a novel sequence-based paradigm that enables scalable RTL timing prediction via linearizing logic cones by breadth-first traversal and applying modern linear sequence models. Furthermore, sequence models are customized by four synergistic techniques, including sequence shuffling, bidirectional modeling, differentiable modeling, and a hybrid graph-sequence architecture. Extensive experiments demonstrate significant improvements of RTL-Sequencer over state-of-the-art baselines, advancing early-stage timing optimization.
Authors: Suzan Awinat, Alfonso Ortega del Puente
Abstract: Metaphor in Arabic is a culturally grounded mechanism for constructing meaning, encoding cultural knowledge that shapes interpretation. Yet current Arabic language models typically collapse lexical, cultural, and metaphorical information into a single representational space, a phenomenon we term "semantic smearing". We introduce CAMMAR (Culture-Aware Matryoshka for Metaphorical Arabic Representations), a representation learning framework that organizes meaning into nested lexical, cultural, and metaphorical embedding subspaces through a staged semantic curriculum. The design implements compositional principles of Al-Jurjani's theory of nazum, modeling figurative meaning as compositionally grounded in prior semantic relations, and yields a training-free geometric measure of metaphoricity based on the distance between lexical and metaphorical representations. Evaluated on a new span-annotated Arabic metaphor set as word-matched figurative/literal pairs, the geometric readout detects metaphor well above chance when the inter-layer geometry is shaped by paired supervision (AUC up to 0.84; figurative outscores its literal counterpart for the same word in 82.6\% of pairs), but sits at chance under an unsupervised domain contrast alone, a clean separation between a legible-under-supervision regime and a non-emergent one. A controlled ablation shows that grounding the lexical layer in morphological roots gives a small but consistent gain, an effect absent from direct probing that reflects the layer's quality as a measurement anchor. We will release the datasets, cultural concept inventory, and code upon acceptance.
Authors: Dimitrios Karageorgiou, Symeon Papadopoulos, Ioannis Kompatsiaris, Efstratios Gavves
Abstract: Autoregressive video diffusion models have enabled the generation of arbitrarily long videos by removing conditioning on future frames, thus greatly improving computational efficiency. Yet, they suffer from error accumulation over time, as the denoised sequence gradually drifts away from the conditioning distribution seen during training. Recent advances attempt to reduce this error by anchoring each generated frame to the learned manifold of real ones. However, even when all generated individual frames lie close to the real manifold, there are trajectories which the model lacks sufficient knowledge to continue without exiting it, thus reaching a terminal point. To prevent the model from being trapped in terminal points, we start from the hypothesis that for well-modeled future trajectories the distribution of the predicted noise should match the one of the forward noising process. To enforce such a prior at test time, we introduce Terminal points Avoidance through Noise Guided Optimization (TANGO), which uses the diffusion model as a critic of its own outputs, by predicting one step forward and requiring an isotropic Gaussian noise prediction. We use the deviation from this expected noise distribution to search for an alternative trajectory that does not lead to a terminal point. Our approach achieves a $3.1\%$ absolute improvement on VBench over state-of-the-art, while reducing Fr\'echet Video Distance by $28.3\%$ on average across $15$s videos. Our code is available on https://mever-team.github.io/tango.
Authors: Muness Castle, Eric Rubeck
Abstract: Coding agents can fix a failing example without preserving the domain rule that made it fail, so later generations can repeat the same plausible mistake. We present agentic synthesis against counterexample-supplemented sketches, a repository-native method for systems whose governing policy is discovered during implementation. A human starts with a partial, code-shaped sketch, and a coding agent generates the first implementation. When a concrete failure exposes missing or mistaken policy, an operator explicitly approves the corrected behavior and rule. The agent then revises the sketch and repairs or regenerates code and prompt surfaces for that one counterexample. The full archive preserves provenance; a selected regression set gates each revision before the next candidate is revealed; and periodic clean regeneration tests whether the evolved sketch, rather than prompt history or accumulated examples, carries the learned policy. We demonstrate the method with CatSynth, a synthetic browser application and captured coding-agent experiment. In one open-world run with GPT-5.4-mini, 8 of 14 frozen candidate cases became counterexamples. The rebuild controls inherited that promotion schedule, and all three paths passed the 8 accepted cases. Rebuilding from the evolved sketch passed 19 of 21 withheld cases, compared with 15 of 21 when rebuilding from the initial sketch and replaying all accepted examples. Retaining code across counterexamples required 9 Developer calls and 719 lines of cumulative artifact churn, versus 15 calls and 2,394 lines for replay-all, and passed 18 of 21 withheld cases. These results provide inspectable evidence that the evolved sketch carried reviewed policy and that retaining code reduced rework in this run; with one model and one reveal order, they do not establish general superiority or correctness beyond the encoded checks.
Authors: Indraveni Chebolu, Rohan Singh, Arnab Mallick, Harmesh Rana
Abstract: Moderation systems increasingly rely on external toxicity tools, but those tools are unreliable under code-mixing, transliteration, slang, and language mismatch. We study the \emph{conditional reliability} of toxicity priors in Indian multilingual and code-mixed short text: English toxicity, Indic abuse, and rule-based severity cues can be useful evidence, but only in some linguistic and abuse-severity contexts. We propose ToxGate, a trust-fusion head that conditions each auxiliary signal on the encoder representation before adding it to the prediction state. Across three short-text abuse datasets, four transformer encoders, and five seeds per setting, ToxGate improves over matched plain encoders in 10 of 12 in-domain settings and 7 of 8 transfer settings. The largest and most interpretable gains occur in high-risk moderation slices, including explicit slurs, violent threats, and cross-dataset transfer. The broader lesson is that moderation systems should treat external toxicity tools and priors as conditional evidence rather than fixed features or ground truth, in focused ablations, source-specific gating gives the strongest results in transfer, severe-abuse slices, and high-risk triage.
Authors: Jiaxi Jiang, Bharat Lal Bhatnagar, Nan Yang, Lingni Ma, Sebastian Starke, Robin Kips, Nadine Bertsch, Christian Holz, Federica Bogo
Abstract: Human motion capture from head-mounted devices (HMDs) offers a scalable way to acquire real-world human motion and interaction data, which is crucial for applications in embodied AI and VR/AR. Existing approaches focus on either egocentric body tracking, estimating the motion of the subject wearing the device, or exocentric tracking, capturing the movements of people in the wearer's surroundings. So far, these two paradigms have largely been explored in isolation. In this paper, we propose a novel distributed framework that jointly leverages ego- and exocentric multi-modal signals for human motion estimation from HMDs. Unlike traditional motion capture systems requiring bulky multi-camera setups or obtrusive mocap suits, our approach, EgoExoMoCap, is as simple as two (or more) people, each wearing a pair of smart glasses. The method leverages head (plus potentially wrist) tracking signals for accurate estimation of global motion in the 3D world and combines context-aware image features based on DINOv3 to achieve robustness in the presence of noise and occlusions. Extensive experiments on two in-the-wild datasets show that our approach can robustly reconstruct motion even in challenging scenarios.
Authors: Jens Frankenreiter
Abstract: Much empirical legal research depends on translating unstructured text into structured variables. In corporate governance research as elsewhere, this translation has traditionally relied on human coding of documents such as charters and bylaws, a process that is costly, difficult to scale, and often opaque. This paper introduces DECODEM, a set of benchmark datasets for evaluating the automated extraction of corporate governance variables from organizational documents. The benchmarks pair randomly sampled corporate charters and bylaws with high-quality human annotations covering a range of governance provisions commonly studied in empirical work. Using these datasets, the paper evaluates several large-language-model extraction pipelines that vary in prompt design, task decomposition, and document handling. The underlying task consists of a set of document-level binary classification problems, one for each governance variable. The results show that automated extraction is feasible at a high level of accuracy for many provisions, with median performance near the upper bound across approaches. At the same time, performance varies systematically across variables, with a small number of provisions accounting for most of the remaining errors. More elaborate prompting strategies and cascading pipelines do not consistently improve performance for frontier models, but substantially narrow the gap between frontier and efficiency-oriented models in some settings, suggesting that pipeline design can partly substitute for model capability. By providing a standardized benchmark and a systematic evaluation of extraction methods, the paper demonstrates that current frontier models can extract legally meaningful information from complex corporate documents with high accuracy and suggests an important future role for automated feature extraction in constructing corporate governance datasets.
Authors: Sebastian Cochinescu
Abstract: Large language models are broadly capable, yet in sustained one-to-one conversation they still read as flat: competent, responsive, and somehow not quite the presence of a mind. We hypothesize that a central missing ingredient is not more capability but dimensional completeness. We propose that the believability of an artificial interlocutor -- the degree to which a user attributes an inner life to it, which we call perceived mind -- is governed by whether the agent expresses a small set of first-person stances that humans use as evidence of mind, and that this is separable from task intelligence. We name four such dimensions -- time, truth, entropy, and love -- each defined as a behavioral stance rather than a benchmark competency, each with a human analog and a concrete emulation path; the time dimension already has an author-reported prototype. We identify an observable behavior layer -- initiative (unprompted action) and cadence (the shape and timing of turns) -- through which the stances surface in conversation, both partially realized as deployed features in a production companion application. We state six falsifiable predictions that a later pre-registered study will test, separating those that are pre-registrable now from those that remain conjectures pending operationalization. This is a conceptual framework: it reports no human-subjects data, and its central comparative claims are predictions, not findings. Throughout we hold a firm boundary -- the object is inferrable interiority, not interiority; this is perception engineering, not a theory of machine consciousness -- and we treat the resulting attachment and manipulation risks as load-bearing rather than incidental.
Authors: Andy Catruna, Emilian Radoi
Abstract: While the internal mechanisms of autoregressive (AR) transformers have been studied extensively, much less is known about diffusion language models (DLMs), an emerging alternative that generates text by iterative denoising. In this work, we study how DLMs implement induction, a mechanism behind in-context learning in which the model finds a repeated context and copies the token that followed it. Our analysis compares attention-only AR models and absorbing-mask DLMs with matched architectures. We find that DLMs learn a bidirectional induction circuit, where previous-token and next-token heads write local context into the residual stream and later induction heads use it to find and copy the answer from the matching source position. The circuit is direction-symmetric, working whether the source appears in the past or in the future. When only left context is visible, matching what an AR model sees, the DLM does not outperform its AR counterpart in induction capabilities. However, we observe it has stronger induction when both sides of the masked token are visible, pointing to bidirectional context access rather than a stronger one-sided mechanism. Beyond induction, we provide causal evidence that DLMs compute the global fraction of masked tokens and use it as an implicit timestep, even though they are given no explicit timestep embedding.
Authors: Sudhanshu Mittal, Arian Mousakhan, Silvio Galesso, Karim Farid, Jonannes Dienert, Rajat Sahay, Thomas Brox
Abstract: Current world models operate at a single level of abstraction, with most prioritizing perceptual fidelity while lacking the spatial reasoning and semantic understanding required for real-world downstream tasks. We present a hierarchical driving world model that factorizes future prediction across two levels operating at distinct temporal and abstraction scales: a high-level predictor that forecasts coarse scene structure over extended temporal horizons, and a low-level generator that produces detailed predictions conditioned on the high-level output. This decomposition yields high perceptual fidelity while also capturing strong spatial and semantic representations. We further show that pretraining with a diffusion forcing objective yields substantially richer internal representations than the standard teacher forcing objective, while teacher forcing -- predicting only the next frame from clean context -- produces more stable autoregressive rollouts. We therefore introduce a generic two-stage training paradigm that pretrains the model with diffusion forcing and fine-tunes with teacher forcing, combining the representational benefits of the former with the rollout stability of the latter. Our approach achieves state-of-the-art results across the standard suite of driving world model evaluations on established benchmarks, including long-horizon generation fidelity, steering responsiveness evaluated on counterfactual scenarios, and internal representation quality. Project page with code, demo, checkpoints and qualitative results: https://lmb-freiburg.github.io/orbis2.github.io/
Authors: Mohamed Amine Kina
Abstract: Data-free knowledge distillation transfers the knowledge encoded in a teacher model to a student model without access to the original training data. Prior work such as Contrastive Abductive Knowledge Extraction (CAKE) achieves this for classifiers by synthesizing samples near the teacher's decision boundary. In this work, we investigate whether this boundary-seeking principle extends to autoencoder distillation through experiments on the MNIST dataset . To enable a direct comparison, we reformulate continuous reconstruction as a dense, per-feature classification task, allowing the decoder to output categorical logits. We show that boundary-seeking objectives are fundamentally ill-posed in bottlenecked generative architectures. CAKE operates on a single, instance-level objective, but a decoder acts as an array of tightly coupled, feature-level classifiers constrained by a shared low-dimensional bottleneck. Independently sampling contrastive targets for these coupled outputs violates the geometry of the learned latent manifold and produces severe gradient conflicts instead of informative boundary samples. Manifold-aware synthesis bypasses these conflicts entirely and establishes an effective baseline for data-free generative distillation.
Authors: Jose Manuel de la Chica Rodriguez, Jairo Rodriguez Arias, Spyridon Chouliaras
Abstract: Standard automation ROI misses four categories of systemic risk -- tacit knowledge erosion, resilience reduction, regulatory exposure, and socio-institutional capital degradation -- that affect long-term organizational performance. PHP-AIO (Protocol for Human Preservation in AI-Optimized Organizations) is a five-gate sequential decision protocol with a final composite check that quantifies these unpriced systemic risks at the role level and produces auditable automation decisions. A closed-form automation-debt measure ($\rho(P)$) formalises how role-level decisions accumulate across multi-step processes; its warning is neutralised only by a regulator-mandated human-in-the-loop anchor. Applied to stylised profiles of representative internal roles, PHP-AIO produces distinct outcomes -- automate, augment, hybrid, and preserve -- for candidates that standard cost-benefit analysis would uniformly automate. Threshold sensitivity analysis confirms the gate decisions are robust to upward perturbations of at least 14% in three of four representative cases. Keywords: AI governance, automation decision, human oversight, tacit knowledge, organizational resilience, financial services
Authors: Elpida Tzafestas
Abstract: We study a society of agents belonging to a number of occupational or cultural groups that form opinions about others' situation in the same or different group. Opinions develop either by observation within own group or by directly interacting with members of other groups, therefore by word of mouth (WoM). Additionally, global mass media (MM) may be available that inform indirectly about the situation of the various groups. The sociocultural interplay of these processes and the degrees of relative exposure to each of the sources has diversified effects on final opinion and social attitude formation. In large and complex societies and groups where not everyone can physically interact by WoM with everyone about everything, these processes show potential for mass control and social automation engineering. Our model can also represent and be generally informative about segmented societies that consist of groups with different occupational and cultural characteristics and it can offer insights into social issues such as the generation gap, social cleavages and so on.
Authors: Josef Hoppe, Sarra Bouchkati, Farah Nasr, Jonathan Krapp, Alexander Och, Maximilian Wirth, Jan Schiefelbein-Lach, Oliver Pohl, Andreas Ulbig, Michael T. Schaub
Abstract: Increases in photovoltaic generation, charging of electric vehicles and heat-pump demand challenge operating limits in low-voltage distribution grids. This requires curative curtailment methods that can operate under sparse observability, noisy measurements, and imperfect grid models. Unlike prior end-to-end reinforcement-learning approaches for partially observable curtailment, this work decouples congestion detection and control by combining a random-forest violation pre-classifier with an actor-critic controller, and evaluates its robustness to measurement noise and grid-parameter mismatch. The framework is tested on a real low-voltage grid using synthetic future operating scenarios with low observability and controllability. With accurate grid parameters, the controller reduces total violation magnitude by 98.9%, and this performance remains nearly unchanged under the tested measurement-noise settings. Grid-model mismatch proves to be more challenging, but the controller still mitigates most violations under the tested mismatch assumptions.
Authors: Jehun Kang, Jungha Wang, Youngjun Hwang, David Hyunchul Shim
Abstract: Multi-Task Learning (MTL) in robotics perception systems supports comprehensive 3D spatial scene understanding by integrating semantic segmentation and depth estimation. While Vision Foundation Models (VFMs) are increasingly adopted as robust feature encoders, existing decoding strategies present a critical bottleneck. To address this, we propose DPNeXt, a streamlined multi-scale feature fusion decoder and efficient alternative to the standard Dense Prediction Transformer (DPT). DPNeXt uses dual depthwise separable inverted bottlenecks to improve frozen VFM utilization through fusion-centric decoding and independent task modularization. To further mitigate negative inductive transfer between tasks, we introduce the Multi-Task Boundary Guidance (MTBG) strategy. Unlike prior boundary-aware methods that add fusion modules or gating, MTBG applies symmetric boundary-focused supervision to encourage geometric consistency without extra annotation or inference cost. Experiments on Cityscapes show that DPNeXt-S outperforms prior state-of-the-art (SOTA) MTL models, while DPNeXt-B further improves the overall performance and achieves the best results among the compared methods. On NYUv2, DPNeXt-B also achieves the best semantic segmentation and depth estimation results among the compared methods while requiring substantially fewer trainable parameters than prior large-scale MTL models. Compared with the standard DPT, DPNeXt-S reduces trainable parameters by 78.6% and achieves the fastest inference speed among the compared models on resource-constrained laptop hardware. The source code, model checkpoints, and a demo video will be made available at https://github.com/kangjehun/DPNeXt.
Authors: Junyuan Zheng, Onkar Salvi, John Chan
Abstract: Dialogue state tracking (DST) is one of the core components in task-oriented dialogue systems. At each turn in a conversation, DST estimates the user belief or dialogue state, which is used as input for downstream modules to predict system actions and generate responses. The increasingly popular dialogue system applications like Google Assistant, Siri and Alexa need to support a large number of services and APIs, resulting in growing attention to the scalability of such systems. Especially for some domains with little or no training data, the capability of transferring existing knowledge of other domains is highly desired. In this paper, we present a novel scalable framework for multi-domain dialogue state tracking. The proposed system leverages the pretrained BERT model to achieve zero-shot generalization, making it easy to quickly adapt to new domains without additional training. The performance of our model is evaluated on recently released schema-based dialogue (SGD) dataset, showing significant improvement compared to previous baseline.
Authors: Yu-Chao Hsu, Yu-Cheng Lin, Tai-Yue Li, Nan-Yow Chen, En-Jui Kuo
Abstract: Quantum continual learning aims to train quantum models on sequential tasks without losing previously learned knowledge. However, variational quantum classifiers (VQCs) are prone to catastrophic forgetting under nonstationary task distributions. We propose quantum elastic weight consolidation (QEWC), a quantum Fisher information (QFI)-informed regularization method for mitigating forgetting. Unlike conventional elastic weight consolidation based on classical Fisher information (CFI), which measures parameter importance through measurement-dependent output statistics, QEWC uses QFI to quantify the intrinsic sensitivity of the parameterized quantum state. This gives an information-geometric view in which important parameters are identified by the local response of the quantum state manifold. We evaluate QEWC on VQCs trained on sequential binary classification tasks, including classical image-classification and quantum phase-classification tasks. Simulations show that sequential training without regularization causes severe forgetting, while both CFI-based EWC and QFI-based QEWC improve retention of previous tasks. Mechanistic analyses further show that the two methods impose different regularization geometries: CFI acts selectively on measurement-sensitive directions, whereas QFI imposes a denser state-geometric constraint over parameter space. Under depolarizing noise, CFI values are strongly suppressed by degraded measurement statistics, while QFI preserves a more stable sensitivity structure of the noisy parameterized quantum state. These results establish QEWC as a physically motivated approach for studying and mitigating forgetting in quantum continual learning through quantum-state geometry.
Authors: Olayiwola Arowolo, Maosheng Yang, Jochen Cremer
Abstract: Data-driven pre-fault dynamic security assessment (DSA) rapidly evaluates the dynamic risk of credible contingencies on a power system using machine learning. Existing approaches face two limitations. First, they require a large labelled database for training, with a separate model trained, tuned, and maintained for each contingency in a potentially long list of credible contingencies. Second, the trained models generalize poorly to unseen contingencies. This work addresses the limitations by using a tabular foundation model (TFM) that assesses stability through in-context learning, requiring no retraining or hyperparameter optimization. A single TFM can assess many contingencies at once, removing the need for one model per classifier. We also characterize when the use of electrical distance coordinates (EDC) as continuous features enables generalization of TFM to unseen contingencies and when they do not, demonstrating how a few labelled samples can reliably improve generalization. Through comprehensive case studies on the IEEE 68-bus system, we show that a single TFM attains an average Macro F1 score of about 90% with only 120 labelled samples per contingency, roughly two orders of magnitude fewer than conventionally assumed, without any model retraining or hyperparameter tuning. For new/unseen contingencies, we show that using just 10 labelled samples of the new contingency with EDC encoding matches the best achievable transfer learning oracle model, which requires fully labelled data and is not deployable in practice. Overall, this initial study paves the way towards developing and deploying foundation models for power system operations, with possible applications across multiple operational tasks.
Authors: Zitian Gao, Yilong Chen, Yihao Xiao, Xinyu Yang, Ran Tao, Joey Zhou, Bryan Dai
Abstract: We present Loopie, the most powerful looped Transformer to date. The Loopie series consists of two Mixture-of-Experts (MoE) models: a 20B-parameter model with 2B active parameters and a 6Bparameter model with 0.6B active parameters. Looped Transformers have long faced a challenge: given an N-fold increase in pre-training compute, increasing the parameter count by a factor of N usually outperforms looping a model N times. Loopie addresses this challenge. Extensive ablation studies, including comparisons with a vanilla 30B-A3B model, show that Loopie substantially outperforms vanilla Transformer baselines trained with the same compute budget. Our novel post-training pipeline equips Loopie with strong reasoning abilities. At the 2025 IMO and IPhO, Loopie achieves gold-medal performance without tools.
Authors: Ajay Patel, Kartik Hosanagar, Ramayya Krishnan, Chris Callison-Burch, Karim Lakhani, Mitch Weiss
Abstract: Large language models (LLMs) are improving rapidly as reflected in benchmark scores, yet these AI benchmarks largely test capabilities such as factual recall, narrow question answering, mathematical problem-solving, and coding and agentic tool-use. What remains poorly measured is AI progress on the analytical knowledge work white-collar professionals perform daily, including synthesizing complex information, exercising judgment under uncertainty and incomplete information, applying strategic and adversarial thinking in multi-stakeholder settings, weighing trade-offs, and producing defensible, structured analyses. This gap is even more pronounced for subjective components of such work, where success can be challenging to define. The "case method" form of education practiced by top business schools provides a natural foundation for addressing this measurement gap, and we construct BusinessCaseBench, a benchmark spanning hundreds of questions drawn from business cases across eighteen disciplines, each paired with a grading rubric derived from the expert-written instructor case solution. On BusinessCaseBench, frontier AI models already score highly against instructor rubrics, and capability within one model family improves substantially over two years. These results provide strong evidence that AI performance on this class of work is already high and rapidly improving, with implications for business schools, where case pedagogy trains undergraduates and MBAs in this kind of analytical reasoning, and for entry-level professional roles, where such skills have historically anchored early-career work.
Authors: S. Aaron McClendon
Abstract: Model merging is promoted as a substitute for joint multi-task training, yet in the reinforcement-learning setting this substitution is essentially never tested against the baseline it claims to replace: methods merge independently released agents precisely because a joint model is unavailable. We build the missing comparison. Training difficulty-1 and difficulty-2 Qwen3-8B specialists on the AppWorld agent benchmark with LOOP, we merge them (TIES, RAM+) and pit the result against a jointly trained model on the same data. On task-goal completion, merging matches joint RL -- and every merge variant is statistically indistinguishable. To explain why merge method does not matter here, we measure the geometry of the specialists' task vectors, which carries no task-sampling noise: they are near-orthogonal (cosine 0.06 - 0.10) despite ~65% support overlap, a small, shared direction that grows over training and that we calibrate against a random-init floor and a same-run ceiling to confirm it reflects learning, not the low-rank parameterization. Because direction and support are decoupled, support and sign-based merging (RAM, TIES) collapse to near-uniform averaging. We release all code and statistics.
Authors: Iker Moran-Cavero, Monica Hernandez, Elvira Mayordomo, Naiara Artiaga, Beatriz Pardi\~nas, Beatriz Cordon, Elena Garcia-Martin
Abstract: Retinal layer segmentation in Optical Coherence Tomography (OCT) is a fundamental step for extracting quantitative biomarkers of retinal structure. Indeed, there is a growing interest in the analysis of OCTs in the context of neurodegenerative diseases. However, segmentation remains challenging due to speckle noise, shadowing artifacts, low contrast between adjacent layers, anatomical variability across subjects, and domain shifts arising from different acquisition protocols and clinical populations. While deep learning methods have achieved remarkable performance, their robustness and generalization across heterogeneous datasets remain limited. In this work, we investigate the role of spatial normalization as a preprocessing strategy to mitigate geometric domain shifts and improve the consistency of retinal layer segmentation. Inspired by standard practices in neuroimaging, we introduce a fovea-centered normalization framework that aligns OCT volumes into a common anatomical reference. We perform a comprehensive evaluation of state-of-the-art deep learning architectures. To provide a comprehensive assessment of segmentation quality, we combine conventional overlap-based metrics at B-scan level with topology-aware metrics at A-scan level and thickness-based measures at the en-face level. In cases where a ground truth is not available, we propose topology violation quantitative metrics that do not require ground truth annotations and a thickness-based qualitative assessment that captures structural consistency and clinically relevant patterns at the en-face level. The results demonstrate the importance of spatial normalization in OCT segmentation pipelines toward the development of robust and clinically meaningful retinal analysis tools, enabling reliable biomarker extraction and downstream computational analysis in neurodegenerative research.
Authors: Mazene Ameur, Abdelkader Mekrache, Bouziane Brik, Adlen Ksentini
Abstract: Agentic Artificial Intelligence (AI), enabled by Large Language Models, marks a shift from rule-based automation toward autonomous, goal-driven control of Next-Generation Networks (NGNs). Existing surveys treat the two domains in isolation, leaving protocol integration, evaluation, and standardization alignment underexplored. To address this gap, a two-part tutorial-and-survey is presented. Part I formalises the control, management, and AI-native planes of 5G and 6G. It then covers the foundations of agentic systems: reasoning, planning, tool use, multi-agent coordination, and evaluation. Part II maps agentic capabilities onto 5G/6G control surfaces, standardization, and major 6G initiatives. Finally, it identifies open challenges shaping autonomous telecommunications.
Authors: Haoran Sun, Wentao Zhang, Junyang Hua, Hedan Yang, Yongjian Guo, Yifei Zhang, Xiaolong Xiang, Mingxi Luo, Jing Long, Chen Zhao, Chen Zhou, Wanting Xu, Qiming Yang, Hui Zhang, Song Wang, Xiaodong Bai, Shuai Di, Xu Chu, Xiaotie Deng, Yicheng Gong, Junwu Xiong
Abstract: The post-training of Vision-Language-Action (VLA) models is essential due to the diversity of simulators, robot embodiments, and task objectives. Existing compute services, whether offered as direct accelerator rental or batch-workload submission, typically allocate an exclusive set of GPU and CPU resources to a single tenant. While this paradigm maximizes client flexibility, it burdens users with infrastructure adaptation, and the fixed card-hour accounting model renders short or bursty workloads both expensive for tenants and inefficient for the service provider. To address these challenges, we present JoyNexus, a unified service for multi-tenant VLA supervised fine-tuning, reinforcement learning, and evaluation. JoyNexus decouples the Training Model Service, Inference Model Service, and Environment Service, each accessed through APIs and backed by resident shared base models with tenant-specific slots. Tenants can directly invoke high-level semantic APIs for training, rollout, and evaluation, or compose custom algorithms using lower-level APIs and their assigned endpoints. Multiple tenants submit workloads concurrently; their action modules, optimizers, rollout records, and policy versions remain isolated, and the service is scheduled by the global Training Queue and Inference Queue. To further improve multi-tenant training efficiency, JoyNexus introduces group batching for heterogeneous VLA data schemas that share a compatible model-facing prefix, enabling a single shared backbone forward pass over grouped samples. Finally, we evaluate JoyNexus through workload simulation and a group-batching pipeline in a realistic embodied scenario. Results show that, compared with isolated single-tenant execution, JoyNexus reduces aggregate GPU time and improves service utilization via cross-tenant scheduling on shared resources.
Authors: Bhavana Verma, Priyanka Meel, Dinesh Kumar Vishwakarma
Abstract: Multimodal sarcasm and cyberbullying detection remain challenging because the intended meaning often emerges from incongruity between textual and visual information rather than from either modality alone. Existing multimodal approaches primarily rely on feature fusion or cross-modal attention, which may not effectively capture hierarchical semantic inconsistencies across different levels of representation. To address this limitation, this paper proposes HCIG (Hierarchical Cross-modal Incongruity Graph Network), a novel framework that models cross-modal incongruity at token, phrase, and global levels using graph attention networks and adaptively integrates these representations through a learned hierarchical attention mechanism. As a complementary architecture, we also introduce GCCN (Graph-based Cross-modal Contradiction Network), which performs graph-based reasoning using contradiction-aware pooling for efficient multimodal interaction learning. The proposed models are evaluated on the MMSD sarcasm benchmark and the MultiBully cyberbullying dataset, together with comprehensive ablation studies and cross-task transfer experiments. Experimental results demonstrate that HCIG achieves the best performance on MMSD with 85.74% accuracy and 85.29% macro-F1, while GCCN attains the highest macro-F1 (68.66%) on MultiBully and HCIG achieves the highest accuracy (69.62%) and bullying-class F1 (74.90%). The findings demonstrate that hierarchical multi-granularity incongruity modeling provides more effective multimodal reasoning than conventional fusion strategies, offering a robust framework for sarcasm and cyberbullying detection in social media.
Authors: Hanyang Chen, Anirudh Satheesh, Longchao Da, Hua Wei
Abstract: Transferring policies across domains poses a vital challenge in reinforcement learning, due to the dynamics mismatch between the source and target domains. In this paper, we consider the setting of online dynamics adaptation, where policies are trained in the source domain with sufficient data, while only limited interactions with the target domain are allowed. There are a few existing works that address the dynamics mismatch by employing domain classifiers, value-guided data filtering, or representation learning. Instead, we study the domain adaptation problem from a generative modeling perspective. Specifically, we introduce DADiff, a diffusion-based framework that leverages the discrepancy between source and target domain generative trajectories in the generation process of the next state to estimate the dynamics mismatch. Both reward modification and data selection variants are developed to adapt the policy to the target domain. We also provide a theoretical analysis to show that the performance difference of a given policy between the two domains is bounded by the generative trajectory deviation. More discussions on the applicability of the variants and the connection between our theoretical analysis and the prior work are further provided. We conduct extensive experiments in environments with various shifts to validate the effectiveness of our method. The results demonstrate that our method provides superior performance compared to existing approaches, effectively addressing the dynamics mismatch. We provide the code of our method at https://github.com/hanyang-chen/DADiff-release
Authors: Jingyan Shen, Ang Li, Salman Rahman, Yifan Sun, Micah Goldblum, Matus Telgarsky, Pavel Izmailov
Abstract: Reinforcement learning (RL) has become central to improving large language models (LLMs) on complex reasoning tasks, yet RL post-training is largely studied in isolation from the pretraining that precedes it. As a result, two basic questions remain open: (1) how do pretraining choices (model size, data) shape the returns to RL compute, and (2) what does RL actually do to the model? These questions are difficult to study in the standard LLM setting: pretraining corpora are vast and uncontrolled, making it hard to attribute behaviors to pretraining versus RL, and systematic compute sweeps across both stages are prohibitively expensive. To address these challenges, we use chess as a controlled testbed for studying reasoning across the full pretraining-to-post-training pipeline. We follow the standard LLM training pipeline by pretraining language models from 5M to 1B parameters on human chess games, supervised fine-tuning on synthetic reasoning traces, and running RL on chess puzzles with verifiable rewards. Using this framework, we find that the post-RL performance at given RL compute level is well-predicted from the pretraining loss, and slope of the RL reward curves improves approximately linearly with the pretraining tokens. Beyond scaling, we find that RL does not simply sharpen the SFT policy: on easy puzzles it amplifies correct moves the SFT policy already preferred, while on hard puzzles it surfaces correct moves that were nearly absent under SFT. We further test whether our findings transfer beyond chess by training a 1B language model on math-domain text, where the same predictive pattern emerges: longer-pretrained checkpoints reach higher post-RL performance and improve faster under RL. In sum, we provide a quantitative account of the pretraining-to-RL interface and a controlled testbed for studying the science of reasoning across the full pretraining-to-post-training pipeline.
Authors: Andrea Ferrario
Abstract: AI governance increasingly requires judgments about whether an AI system remains adequately trustworthy over time, whether observed changes are tolerable, and how such judgments should be documented in a transparent and contestable way. Yet existing work on AI trustworthiness remains either too high-level to support lifecycle monitoring and reassessment or too narrowly metric-driven to connect with governance needs. We therefore propose a lightweight methodology for auditable trustworthiness levels in AI governance. The methodology has two components: a formal framework for representing and learning trustworthiness levels, and a lightweight AI lifecycle governance procedure for documenting, monitoring, and reassessing them over time. The formal framework models governance-relative trustworthiness through a context-sensitive protocol of measurable dimensions and learns trustworthiness levels as interpretable rules over trustworthiness profiles. Using decision trees as an interpretable proof-of-concept model class, the methodology yields explicit trustworthiness plateaus, readable level transitions, and two simple lifecycle diagnostics: boundary margins and profile drift. The governance procedure embeds these formal objects in a conformity-oriented workflow for design-time labeling, post-deployment monitoring, reassessment, and reporting. It also assigns human responsibilities and control gates for protocol design, validation, monitoring, and reassessment. We illustrate the methodology on synthetic AI lifecycle traces involving degradation, shocks, updates, heterogeneous monitoring cadences, and system comparison. Our methodology does not replace legal or other expert judgment: it supports conformity documentation and lifecycle monitoring by providing an evidential basis for documenting and tracking AI governance-relevant changes over time.
Authors: Binglin Zhou, Peng Shi, Ryo Kamoi, Nan Zhang, Rui Zhang
Abstract: Multimodal Scientific Claim Verification (MSCV) requires models to verify scientific claims using visually grounded evidence from papers, including figures, tables, charts, and textual context. However, existing methods often fail because they struggle to locate decisive visual evidence, accurately read structured scientific visuals, and integrate multimodal observations into reliable reasoning. We introduce ToolSciVer, the first tool-augmented framework for MSCV to our knowledge. ToolSciVer equips a VLM with three type-aware visual tools, table row/column focus, chart-to-structure parsing, and high-resolution region zoom, which convert dense scientific visuals into explicit, claim-facing evidence, and trains the policy with Group Relative Policy Optimization (GRPO) under a composite reward of answer correctness, format validity, length control, tool-use efficiency, and tool-validity penalties. Experiments on SciVer and MuSciClaims datasets on five VLMs from three model families (Qwen, InternVL, Gemma) demonstrate that our method achieves superior performance compared to four competitive baselines including prompting-based and RL-based tool-use methods, highlighting the effectiveness of learned, type-aware tool use for scientific claim verification.
Authors: Wendi Yu, Lianhao Zhou, Xiangjue Dong, Sai Sudarshan Barath, Declan Staunton, Byung-Jun Yoon, Xiaoning Qian, James Caverlee, Shuiwang Ji
Abstract: LLM powered multi-agent systems (MAS) have emerged as a promising paradigm for complex tasks. However, their advantages over single-agent systems (SAS) remain unclear, with performance varying inconsistently across settings. Here, we provide an information bottleneck perspective on elucidating the differences between MAS and SAS. Specifically, our key observation is that a SAS accumulates its full reasoning trace in one shared context, while a MAS uses isolated local contexts connected by bounded relay messages. We show that, under infinite relay bandwidth, any SAS can be simulated by a MAS that transmits the full upstream context. Thus, the nontrivial advantage of MAS arises under bounded relays, where compression introduces a fundamental trade-off: reducing redundant context can improve efficiency, but may also incur loss of task-relevant information. We formalize this trade-off as an information bottleneck controlled by an effective parameter $\beta$, which captures how the balance shifts with model capability, and shows that MAS gains arise when context reduction outweighs relay information loss. We conduct 18 controlled experiments across five benchmarks and three model scales to validate our theoretical studies. We observe that MAS consistently helps when relays are near-sufficient, especially for weaker models. In contrast, MAS gains shrink or reverse when relays incur information loss, especially for stronger models that can already extract useful information from redundant context and thus gain little from compression. Our study shows that multi-agent design is fundamentally an information-bottleneck optimization problem. This perspective explains when bounded inter-agent communication helps or hurts.
Authors: Jiarui Zhang, Muzi Tao, Shangshang Wang, Ollie Liu, Xuezhe Ma, Willie Neiswanger
Abstract: Human vision is a closed loop: gaze is continuously redirected by intermediate hypotheses rather than a single snapshot. Decades of psychophysics and cognitive science have argued that this active observation is essential for a wide range of tasks. Whether today's multimodal large language models (MLLMs) exercise active observation is an empirical question that current vision-language benchmarks do not answer. We introduce ActiveVision, a benchmark that makes active observation measurable for MLLMs, comprising 17 tasks across 3 categories. Tasks are designed to force repeated visual perception rather than a single static description. Frontier MLLMs collapse on ActiveVision: the highest-scoring model we evaluate, GPT-5.5 at the highest exposed reasoning-effort tier, solves only 10.6% of items and scores zero on 11 of the 17 tasks, and even Claude Fable 5, despite topping most reasoning and coding leaderboards, solves just 3.5%, far behind three human participants who average 96.1%. Furthermore, much of the gap persists even when models write and run their own vision code: such code is unreliable on realistic imagery, and catching its failures itself requires the active perception the models lack. Together, these results indicate that current MLLMs lack robust active visual observation, motivating architectures and training objectives that close the perception-reasoning loop.
Authors: Kai Ruan, Jinghao Lin, Zihe Huang, Ziqi Zhou, Qianshan Wei, Xuan Wang, Hao Sun
Abstract: Muon is competitive with AdamW in large-scale pre-training, but its value for reinforcement-learning (RL) post-training remains unclear. We study vanilla Muon in sparse-reward agentic RL through matched single-seed comparisons with AdamW on ALFWorld using Qwen2.5-0.5B-Instruct. Under Group-in-Group Policy Optimization (GiGPO), applying Muon only to hidden weight matrices raises final-window validation success from 0.290 to 0.546 (+88%); high-rate AdamW controls retain no post-update success. The effect depends on the advantage estimator and learning rate. At 3e-5, Muon improves GRPO from 0.161 to 0.268, whereas GraphGPO's late-window gap narrows near saturation. At 1e-5, GraphGPO Muon reaches 0.901, raises normalized validation AUC from 0.399 to 0.556, and reaches 0.5 and 0.75 success 30 and 60 updates earlier, respectively. These exploratory results show that Muon can benefit agentic RL and motivate studying the policy optimizer, advantage estimator, and learning rate jointly. Multi-seed and cross-task validation remain open.
Authors: Md Erfan, Ahmed Ryan, Md Kamal Hossain Chowdhury, Md Rayhanur Rahman
Abstract: Connected and Autonomous Vehicles (CAVs) rely on interconnected software and hardware components, including sensors, Electronic Control Units, in-vehicle infotainment systems, and telematics units, where vulnerabilities can compromise assets, users, and vehicle operations. These vulnerabilities are commonly documented as plain text in the Common Vulnerabilities and Exposures (CVE) database; however, security practitioners require structured information about affected assets, types of weaknesses, and attack behaviors to effectively mitigate the risks from these vulnerabilities. To this end, we evaluate open-weight Large Language Models (LLMs) for generating Structured Threat Information Expression (STIX), a well-known structured format for representing threat information, for CAV-related CVEs. We construct a dataset called CAV-STIXGen that maps CAV vulnerability descriptions to STIX domain objects (SDO), STIX relationship objects (SRO), Common Weakness Enumeration (CWE), and MITRE ATT&CK techniques mappings. Using this dataset, we evaluated 11 open-weight LLMs (4B to 120B parameters) across various prompting strategies and temperatures. Single-model configurations achieve F1 scores of 0.94 for SDO, 0.63 for SRO, and 0.99 for CWE mapping, while complete MITRE ATT&CK mapping remains challenging. In a multi-agent setup, Gemma-4-31B and Codestral-22B achieve F1 scores of 0.91 for SDOs and 0.43 for SROs, respectively. Lastly, we analyze CWE and MITRE ATT&CK co-occurrences to identify recurring threat patterns in the CAV domain, demonstrating how AI-assisted vulnerability-to-STIX translation can automate threat intelligence and prioritize defense in transportation security.
Authors: Lu Guo, Yixiang Shan, Zhengbang Zhu, Qifan Liang, Lichang Song, Ting Long, Weinan Zhang, Yi Chang
Abstract: Offline reinforcement learning (RL) learns policies from fixed datasets, thereby avoiding costly or unsafe environment interactions. However, its reliance on finite static datasets inherently restricts the ability to generalize beyond the training distribution. Prior solutions based on synthetic data augmentation often fail to generalize to unseen scenarios in the (augmented) dataset. To address these challenges, we propose Retrieval High-quAlity Demonstrations (RAD) for decision-making, which innovatively introduces a retrieval mechanism into offline RL. Specifically, RAD retrieves high-return and reachable states from the offline dataset as target states, and leverages a generative model to generate sub-trajectories conditioned on these targets for planning. Since the targets are high-return states, once the agent reaches such a target, it can continue to obtain high returns by following the associated high-return actions, thereby improving policy generalization. Extensive experiments confirm that RAD achieves competitive or superior performance compared to baselines across diverse benchmarks, validating its effectiveness. Our code is available at https://github.com/LeahGL/RAD.
Authors: Raffaele Pojer, Andrea Passerini, Kim G. Larsen, Manfred Jaeger
Abstract: Graph neural networks (GNNs) excel at predictive tasks on graph-structured data but often lack the ability to incorporate symbolic domain knowledge and perform general reasoning. Relational Bayesian Networks (RBNs), in contrast, enable fully generative probabilistic modeling over graph-like structures and support rich symbolic knowledge and probabilistic inference. This paper presents a neuro-symbolic framework that seamlessly integrates GNNs into RBNs, combining the learning strength of GNNs with the flexible reasoning capabilities of RBNs. We develop two implementations of this integration: one compiles GNNs directly into the native RBN language, while the other maintains the GNN as an external component. Both approaches preserve the semantics and computational properties of GNNs while fully aligning with the RBN modeling paradigm. We also propose a maximum a posteriori (MAP) inference method for these neuro-symbolic models. To demonstrate the framework's versatility, we apply it to two distinct problems. First, we transform a GNN for node classification into a collective classification model that explicitly models homo- and heterophilic label patterns, substantially improving accuracy on both synthetic and real-world datasets. Second, we introduce a multi-objective network optimization problem in environmental planning, where MAP inference and continuous numeric relaxation support a complex decision-making task. Both applications include new publicly available benchmark datasets. This work introduces a powerful and coherent neuro-symbolic approach to graph data, bridging learning and reasoning in ways that enable novel applications and improved performance across diverse tasks.
Authors: In-Chang Baek, Seoyoung Lee, Sung-Hyun Kim, Geumhwan Hwang, KyungJoong Kim
Abstract: Human-aligned AI is a critical component of co-creativity, as it enables models to accurately interpret human intent and generate controllable outputs that align with design goals in collaborative content creation. This direction is especially relevant in procedural content generation via reinforcement learning (PCGRL), which is intended to serve as a tool for human designers. However, existing systems often fall short of exhibiting human-centered behavior, limiting the practical utility of AI-driven generation tools in real-world design workflows. In this paper, we propose VIPCGRL (Vision-Instruction PCGRL), a novel deep reinforcement learning framework that incorporates three modalities-text, level, and sketches-to extend control modality and enhance human-likeness. We introduce a shared embedding space trained via quadruple contrastive learning across modalities and human-AI styles, and align the policy using an auxiliary reward based on embedding similarity. Experimental results show that VIPCGRL outperforms existing baselines in human-likeness, as validated by both quantitative metrics and human evaluations. The code and dataset are available at https://github.com/bic4907/VIPCGRL.
Authors: Ruike Hu, Shulei Wu
Abstract: The Structure Gap between probabilistic LLM generation and deterministic schema requirements hinders automated workflows. We propose RL-Struct, a lightweight framework using Gradient Regularized Policy Optimization (GRPO) with a hierarchical reward function to align LLMs with structural constraints. This approach eliminates the critic network, reducing peak VRAM by 38% compared to PPO. On complex JSON tasks, RL-Struct achieves 89.7% structural accuracy and 92.1% validity, significantly outperforming SFT and zero-shot baselines. We also report an emergent curriculum--a self-organized learning process where the model prioritizes syntax before semantics. Our model is publicly available at https://huggingface.co/Freakz3z/Qwen-JSON.
Authors: Bianca Raimondi, Maurizio Gabbrielli
Abstract: The black-box nature of Large Language Models necessitates novel evaluation frameworks that transcend surface-level performance metrics. This study investigates the internal neural representations of cognitive complexity using Bloom's Taxonomy as a hierarchical lens. By analyzing high-dimensional activation vectors from different LLMs, we probe whether different cognitive levels, ranging from basic recall (Remember) to abstract synthesis (Create), are linearly separable within the model's residual streams. Our results demonstrate that linear classifiers achieve approximately 95% mean accuracy across all Bloom levels, providing strong evidence that cognitive level is encoded in a linearly accessible subspace of the model's representations. These findings provide evidence that the model resolves the cognitive difficulty of a prompt early in the forward pass, with representations becoming increasingly separable across layers.
Authors: Katherine Elkins
Abstract: AI development has a fiction dependency problem. Developers have treated large corpora of modern books, including fiction, as valuable enough to accept substantial cost and legal risk, yet current models still struggle to generate compelling long-form fiction. I term this the "AI-Fiction Paradox," and it is particularly startling because training data strongly shapes model output. This paper offers a theoretically precise account of why fiction resists AI generation by identifying three distinct challenges for current systems. First, fiction depends on what I call narrative causation, a form of plot logic where events must feel both surprising in the moment and retrospectively inevitable. Standard autoregressive generation commits to prose sequentially, creating a practical obstacle to coordinating local surprise with retrospective inevitability across a long narrative. Second, I identify an informational revaluation challenge: fiction repeatedly requires the significance of earlier details to be reinterpreted in light of later developments, a form of long-range reasoning that current systems perform unreliably. Third, drawing on over seven years of collaborative research on sentiment arcs, I argue that fiction that moves us requires multi-scale emotional architecture, the orchestration of sentiment at word, sentence, scene, and arc levels simultaneously. Together, these three challenges help explain both why developers have sought large modern book corpora and why compelling long-form fiction remains so difficult to replicate. The analysis also raises urgent questions about what happens when these challenges are overcome. Fiction concentrates unusually powerful cognitive and emotional patterns for modeling human behavior, and mastery of these patterns by AI systems would represent not just a creative achievement but a potent vehicle for human manipulation at scale.
Authors: Kuangshi Ai, Haichao Miao, Kaiyuan Tang, Nathaniel Gorski, Jianxin Sun, Guoxi Liu, Helgi I. Ingolfsson, David Lenz, Hanqi Guo, Hongfeng Yu, Teja Leburu, Michael Molash, Bei Wang, Tom Peterka, Chaoli Wang, Shusen Liu
Abstract: Recent advances in large language models (LLMs) have enabled agentic systems to translate natural-language intent into executable scientific visualization (SciVis) tasks. Despite rapid progress, the community lacks a principled and reproducible benchmark for evaluating these emerging SciVis agents in realistic, multi-step analysis settings. We present SciVisAgentBench, a comprehensive and extensible benchmark for evaluating scientific data analysis and visualization agents. Our benchmark is grounded in a structured taxonomy spanning four dimensions: application domain, data type, complexity level, and visualization operation. It currently comprises 108 expert-crafted cases covering diverse SciVis scenarios. To enable reliable assessment, we introduce a multimodal outcome-centric evaluation pipeline that combines LLM-based judging with deterministic evaluators, including image-based metrics, code checkers, rule-based verifiers, and case-specific evaluators. We also conduct a validity study with 12 SciVis experts to examine the agreement between human and LLM judges. Using this framework, we evaluate representative SciVis agents and general-purpose coding agents to establish initial baselines and reveal capability gaps. SciVisAgentBench is designed as a living benchmark to support systematic comparison, diagnose failure modes, and drive progress in agentic SciVis. The benchmark is available at https://scivisagentbench.github.io/.
Authors: Sophie Chiang, Tom Brennan, Fethiye Irmak Dogan, Jiaee Cheong, Hatice Gunes
Abstract: In recent years, the integration of multimodal machine learning in wellbeing assessment has offered transformative potential for monitoring mental health. However, with the rapid advancement of Vision-Language Models (VLMs), their deployment in clinical settings has raised concerns due to their lack of transparency and potential for bias. While previous research has explored the intersection of fairness and Explainable AI (XAI), its application to VLMs for wellbeing assessment and depression prediction remains under-explored. This work investigates VLM performance across laboratory (AFAR-BSFT) and naturalistic (E-DAIC) datasets, focusing on diagnostic reliability and demographic fairness. Performance varied substantially across environments and architectures; Phi3.5-Vision achieved 80.4% accuracy on E-DAIC, while Qwen2-VL struggled at 33.9%. Additionally, both models demonstrated a tendency to over-predict depression on AFAR-BSFT. Although bias existed across both architectures, Qwen2-VL showed higher gender disparities, while Phi-3.5-Vision exhibited more racial bias. Our XAI intervention framework yielded mixed results; fairness prompting achieved perfect equal opportunity for Qwen2-VL at a severe accuracy cost on E-DAIC. On AFAR-BSFT, explainability-based interventions improved procedural consistency but did not guarantee outcome fairness, sometimes amplifying racial bias. These results highlight a persistent gap between procedural transparency and equitable outcomes. We analyse these findings and consolidate concrete recommendations for addressing them, emphasising that future fairness interventions must jointly optimise predictive accuracy, demographic parity, and cross-domain generalisation.
Authors: Girish Narayanswamy, Maxwell A. Xu, A. Ali Heydari, Samy Abdel-Ghaffar, Marius Guerard, Kara Vaillancourt, Zhihan Zhang, Jake Garrison, Levi Albuquerque, Dimitris Spathis, Hong Yu, Hamid Palangi, Xuhai "Orson" Xu, David G. T. Barrett, Joseph Breda, Jed McGiffin, Yubin Kim, Yuwei Zhang, Naghmeh Rezaei, Samuel Solomon, Karan Ahuja, Tim Althoff, Jake Sunshine, Ming-Zher Poh, Benjamin Yetton, Ari Winbush, Nicholas B. Allen, James M. Rehg, Isaac Galatzer-Levy, Yun Liu, John Hernandez, Anupam Pathak, Conor Heneghan, Yuzhe Yang, Ahmed A. Metwally, Pushmeet Kohli, Mark Malhotra, Shwetak Patel, Xin Liu, Daniel McDuff
Abstract: While ubiquitous wearable sensors capture a wealth of behavioral and physiological information, effectively transforming these signals into personalized health insights is challenging. Specifically, converting low-level sensor data into representations capable of characterizing higher-level states is difficult due to high phenotypic diversity and variation in individual baseline health, physiology, and lifestyle factors. Moreover, collecting wearable data paired with health outcome annotations is laborious and expensive, and retrospective annotation remains practically unfeasible, contributing to a scarcity of data with high-quality labels. To overcome these limitations, we propose a foundation model for wearable health that is pretrained on more than one trillion minutes of unlabeled sensor signals drawn from a large cohort of five million participants. We demonstrate that the joint scaling of model capacity and pretraining data volume leads to systematic improvements in performance, as evaluated on a diverse set of 35 health prediction tasks, spanning cardiovascular, metabolic, sleep, and mental health, as well as lifestyle choices and demographic factors. We find that this population scale representation unlocks label-efficient few-shot learning and generative capabilities for robust daily metric estimation. To further leverage this learned representation, we deploy a classroom of LLM agents to autonomously search the space of downstream predictive heads built on the model embeddings, showing broad performance improvements that increase with LLM model capacity. Finally, we show how integrating these downstream predictors into a Personal Health Agent can support model responses that are more relevant, contextually aware, and safe, and we validate this via 1,860 ratings from a cohort of clinicians.
Authors: Liya Zhu, Jingzhe Ding, Jian Zhang, Jianbo Xue, Shihao Liang, Ge Zhang, Yi Zhu, Duju Zeng, Xiang Gao, Qingshui Gu, Mailun Gao, Huimin Che, Yan Zhao, Peiheng Zhou, Haojun Wang, Chaobo Xian, Lili Le, Chi Wu, Yiwei Liu, Shengda Long, Jiale Yang, Fangzhi Xu, Sijin Wu, Haodong Duan, Chao He, Zhaojian Li, Minchao Wang, Huan Zhou, Jiani Hou, Chuqian Yu, Weiran Shi, Hongwan Gao, Jiamin Chen, Guanhong Chen, Tingqin Luo, Kaiyuan Zhang, Zhixin Yao, Qing Hua, Yuhao Jiang, Jin Chen, Pu Chen, Zhenyu Hu, Xingyu Li, Zhengxuan Jiang, Meng Cao, Tianfeng Long, Haozhe Wang, Mingzhang Wang, Yichen Zhang, Yiming Dai, Chenchen Zhang, Jiaying Wang, Xinying Liu, Xingzu Liu, Lingling Zhang, Xinjie Chen, Yujia Qin, Wangchunshu Zhou, Zhiyong Wu, Yang Liu, Jiaheng Liu, Lei Zhang, Shen Yan, Wenhao Huang, Zaiyuan Wang, Xiaolong Chang
Abstract: Recent years have witnessed the rapid evolution of AI agents toward handling increasingly complex, real-world tasks. However, existing benchmarks rarely evaluate whether agents can operate graphical user interfaces to complete long-horizon, high-value professional workflows across diverse domains. Current GUI benchmarks still predominantly focus on general-purpose software, relatively simple applications, and short-horizon tasks, leaving it largely unknown whether modern agents can follow user instructions to autonomously operate domain-specific professional software and accomplish economically valuable work in an end-to-end manner. To bridge this gap, we introduce Workflow-GYM, a benchmark for long-horizon GUI tasks centered on professional domains and specialized software environments. Through extensive experiments on state-of-the-art models, we find that even the strongest models achieve only slightly above 30% success rates, highlighting that professional long-horizon GUI workflows remain highly challenging for current GUI agents. Further analysis reveals that current agents struggle to maintain long-horizon workflow consistency, frequently exhibiting workflow stage omission, error propagation, objective drift, and insufficient understanding of professional software environments. Our findings provide important insights into the limitations of current agent systems and suggest key directions for the next generation of GUI-agent research.
Authors: Zongsheng Cao, Bihao Zhan, Jinxin Shi, Jiong Wang, Fangchen Yu, Zhijie Zhong, Yingnan Han, Zijie Guo, Tianshuo Peng, Zhuo Liu, Yi Xie, Xiang Zhuang, Shengji Tang, Yue Fan, Runmin Ma, Shiyang Feng, Xiangchao Yan, Anran Liu, Peng Ye, Wenlong Zhang, Xiaosong Wang, Shufei Zhang, Chunfeng Song, Fenghua Ling, Jie Zhou, Liang He, Bo Zhang, Lei Bai
Abstract: Current LLM-based research agents have advanced through agent orchestration, yet largely overlook scientific knowledge orchestration. Existing works often reduce papers to abstracts, surface mentions, and flat \texttt{cites} edges, omitting key entities, claims, evidence, mechanisms, and method lineages essential for scientific reasoning. To this end, we introduce \textbf{Agents-K1}, an end-to-end knowledge orchestration pipeline that converts raw documents into agent-native scientific knowledge graphs. Agents-K1 integrates three components under a unifying theoretical foundation: a multimodal parser whose five-module schema captures entities, multimodal evidence, citations, and typed inter-entity relations across the full paper rather than abstracts alone; a 4B information-extraction backbone trained with GRPO under a rule-based reward; and a graphanything CLI, a tri-source agent interface that unifies web search, multimodal graph retrieval, and cross-document traversal. On top of this, we process 2.46 million scientific papers across six subjects to produce \textbf{Scholar-KG}, of which we release a one-million-paper subset, and the full Scholar-KG is accessible via the SCP link below. The same pipeline can be extended to general-domain corpora and to schema-conformant data synthesis. Extensive experiments demonstrate that Agents-K1 achieves superior performance in scientific information extraction, knowledge graph construction, and multi-hop scientific reasoning.
Authors: Siqi Wang, Daobo Sun, Yizheng Wang, Yilong Zhang, Yabin Jin, Xiaoying Zhuang, Timon Rabczuk
Abstract: Plate and shell structures are widely used in engineering fields. Rapid response prediction for such structures under complex geometries, heterogeneous materials, and varying loads is important for engineering design, but conventional numerical methods usually require repeated modeling and solution when the physical configuration changes. To address this issue, this study proposes a geometry-aware variational physics-informed neural operator (GA-VINO) for Mindlin-Reissner plates. GA-VINO represents the plate geometry using boundary point clouds and incorporates a material encoder, a load encoder, and a scalar-parameter branch to handle spatially random material fields, spatially varying pressure loads, and sample-level uniform parameters. Through multi-branch point cloud encoding and cross-attention, GA-VINO fuses geometric, material, loading, and query point information, and predicts the transverse deflection and rotations at arbitrary query locations. Unlike conventional data-driven neural operators, GA-VINO requires no labeled solution data during training. Instead, it minimizes a variational physics-informed loss constructed from the discretized total potential energy of the Mindlin-Reissner plate. Compared with grid-based neural operators, GA-VINO directly processes irregular point clouds and allows different physical fields to be discretized on different point sets, avoiding forced interpolation onto a common grid. The method is validated on multiple examples involving different geometries, material fields, and load distributions. The results show that GA-VINO achieves promising accuracy in deflection, rotation, gradient-sensitive, and energy-based metrics, completes full-field inference for new samples within milliseconds, and exhibits promising cross-geometry generalization capability.
Authors: Songjun Tu, Chengdong Xu, Qichao Zhang, Yiwen Ma, Yaocheng Zhang, Linjing Li, Dong Li, Xiangyuan Lan, Dongbin Zhao
Abstract: Skill memories can improve agentic reinforcement learning by reusing past experience as textual guidance, but retrieved skills are not oracular: they may help in one state while misleading the same policy in another. This makes the common privileged-teacher assumption fragile, namely that a skill-conditioned prompt can be treated as a fixed teacher for the no-skill prompt. We introduce UCOB, a framework for learning to utilize and evolve agentic skills via credit-aware on-policy bidirectional self-distillation. UCOB treats skill-conditioned and no-skill prompts as two on-policy context views of the same model, compares their return-to-go within the same task and anchor state, and uses the higher-return view as the local teacher. This local credit signal internalizes useful skill-conditioned behavior, corrects misleading skill usage, and guides task/state skill memory updates, utility-aware retrieval, and reflection self-training. Experiments on agentic tasks, including ALFWorld, WebShop, and Search-QA, show that UCOB outperforms skill-free RL, skill-memory baselines, and self-distillation methods across model scales, with up to 23.5 and 18.0 point gains over SOTA baselines on ALFWorld and WebShop. Ablations and analyses further validate its core mechanisms, continual adaptation across environments, and modest training overhead. Code is available at https://github.com/TU2021/UCOB.
Authors: Daiki E. Matsunaga, Junho Na, Tri Wahyu Guntara, Scott Sanner, Pascal Poupart, Jongmin Lee, Kee-Eung Kim
Abstract: Cooperative tasks in Multi-Agent Reinforcement Learning (MARL) require agents to collectively maximize a shared return. Under the Centralized Training with Decentralized Execution (CTDE) paradigm, policy gradients have remained difficult to compute directly. Prior methods largely follow two approaches: independent factorized updates with centralized critics, which lack general joint-improvement guarantees without value decomposition assumptions, or alternating best-response updates, which can converge to suboptimal Nash Equilibria. In this paper, we show the joint policy gradient admits an exact decentralized decomposition of per-agent terms, each formed from per-agent score functions and decentralized critics. Based on this decomposition, we develop Agent-Chained Policy Optimization (ACPO), where actors are trained independently, with their updates together constituting a single step on the joint policy gradient. Central to this result is a serialized view of the simultaneous joint decision in which agents commit actions one at a time, each conditioning on a belief over preceding actions that ties the independent per-agent updates into a single joint step. We evaluate on-policy and off-policy instantiations of ACPO on Multi-Robot Warehouse, SMACv2, and MA-MuJoCo, where it outperforms strong baselines, with the gap widening as the number of agents grows.
Authors: Tom Adamczewski, David Owen, David Rein, Florian Brand, Giles Edkins, Allen Hart, Daniel O'Connell
Abstract: AI models are rapidly improving at autonomous coding, as shown by benchmark progress and one-off demonstrations such as AI implementing a C compiler. However, existing coding benchmarks tend to focus on shorter tasks, and one-off demonstrations are hard to compare systematically because they often have some human guidance, and are not standardized or repeated across models. To address these challenges, we introduce MirrorCode, a long-horizon coding benchmark based on reimplementing entire software projects. In MirrorCode, AI agents must replicate the functionalities of an existing program, without access to its source code. AI solutions must match the original program's output exactly on end-to-end tests, including held-out tests. MirrorCode's 25 target programs span different areas of computing: Unix utilities, data serialization and query tools, bioinformatics, interpreters, static analysis, cryptography, and compression. Existing AI models can already reimplement complex software, with the strongest model scoring 56% across the benchmark. For example, AI can reimplement gotree, a 16,000-line bioinformatics toolkit - a task that we believe would take weeks for a human engineer. However, studying the frontier of performance requires a larger inference budget than typical benchmarks, for example, \$2,600 over 19 days for a single attempt on a large task. We show that AI agents can already complete long-horizon software engineering tasks, especially when requirements are precisely specified. More broadly, our work suggests AI will have transformative effects on software engineering, as autonomous agents continue to improve.
Authors: Bailey Flanigan, Michelle Si
Abstract: Local pairwise comparisons are a standard tool for learning how people want decision rules to work, e.g., in participatory design or alignment. However, their use builds in two strong assumptions: that local comparisons are sufficient evidence about how a person wants an automated decision rule to behave, and that people can always answer those comparisons decisively. We investigate how these assumptions may be compromised under internal pluralism: the idea that an individual evaluates decision rules according to multiple authoritative priorities about how the rule should behave. We provide a formal model of such pluralistic preferences over decision rules, which then lets us identify two distinct failures of forced local pairwise comparison data. First, priorities such as proportionality, egalitarianism, and equal treatment are inherently global: what they imply in one case can depend on what happens elsewhere, so local comparisons may fail to capture them. Second, even when priorities are representable locally, tension between strongly-held priorities can generate internal conflict, producing potentially costly behavioral distortions when comparisons are forced. We then use our model to investigate the alternative -- allowing people to report indecision -- and our findings suggest that doing so can considerably reduce the number of queries needed to learn preferences accurately. We conclude by describing how our model points toward preference-learning methods that elicit these priorities directly, yielding more faithful and interpretable accounts of what people value.
Authors: Andrew Zhang, Chengzhan Li
Abstract: Pooling, substituting, or reusing evaluator-derived step rewards assumes that their direction survives a change of evaluation channel. The same frozen transition can violate that assumption. Process rewards vary agent states, while evaluator audits vary scoring configurations; neither first difference isolates their interaction. We define Agent Step Value (ASV) as a channel-indexed target-margin gain and identify the state-by-channel interaction on complete matched faces. Across frozen PubMed question-answering transitions, direct scoring yields a positive mean ASV, while the generated-view channel yields a negative mean. Two matched replay waves reproduce this reversal, and cross-channel sign disagreement exceeds same-channel retry disagreement by 48.0 percentage points. Matched retrieval faces localize the reversal to the generated-view coordinate and trace its direction across a readout-and-stack bridge. A source-only generation contract restores the positive mean direction on artifact-bearing retrievals and removes parser-detected substantive support claims from artifact-free before-state views. ASV turns channel sensitivity into an identified measurement problem that can be localized and tested by intervention before step rewards are reused.
Authors: Qian Jiang, Zhecheng Shi, Jingpu Yang, Zirui Song, Miao Fang
Abstract: The rapid integration of Large Vision-Language Models (VLMs) into critical infrastructure promises to revolutionize personalized healthcare and dietary management. However, in the domain of food systems, autonomous agents face a unique and persistent challenge: the "Systemic Information Asymmetry" between visual appearance and intrinsic nutritional composition. Existing benchmarks primarily focus on coarse-grained classification tasks, such as food category recognition, which fail to evaluate the intricate reasoning chain required for real-world dietary management -- specifically, the ability to traverse from identifying hidden ingredients to estimating physical mass, and finally synthesizing safety-critical medical advice. In this paper, we introduce OmniFood-Bench, a comprehensive benchmark constructed from the MM-Food-100K dataset. Unlike previous works, OmniFood-Bench evaluates VLMs across three progressive capabilities: Basic Perception (Ingredients & Cooking Methods), Quantitative Reasoning (Portion Size & Nutritional Profiling), and Safety-Critical Advisory (Disease-Specific Recommendations). We evaluate six state-of-the-art VLMs, including gpt-5.1, gemini-3-flash, and qwen3-vl-8B. Our extensive experiments reveal a startling "Semantic-Physical Gap": while models achieve near-human accuracy in naming dishes, they exhibit catastrophic failure in mass estimation and frequently hallucinate benign advice for high-risk diabetic profiles. This work establishes a rigorous standard for trustworthiness in autonomous agents deployed for public health. The code and datasets are available in: https://anonymous.4open.science/r/OmniFood-Bench-7D0B
Authors: Yossi Eliaz
Abstract: Snapshot-backed sandboxes make branching cheap while leaving evidence dependence unchanged. Branches can reuse a model, prompt, repository, tests, observations, or execution ancestor, so counting outputs can amplify one repeated error into high-confidence consensus. We introduce an \emph{evidence-aware reduction contract}: each worker reports an estimate, estimated information, evidence identifiers, fork lineage, and execution metadata. For independent workers estimating one common parameter, we use standard inverse-information pooling in its Gaussian/Wald form. The fixed-dimensional numeric summary can merge in any tree order; evidence IDs and lineage follow separate rules. The residual $\Delta$ measures disagreement, becomes Cochran's $Q$ in the scalar inverse-variance case, and appears in the product integral. A reference implementation validates serialized records, rejects repeated nonempty evidence identifiers, carries evidence and lineage through tree reduction, and uses Cholesky-based numerical linear algebra. Unit tests and seeded synthetic checks exercise the algebra, unequal information, and forged precision; one four-worker named-snapshot trace exercises the end-to-end path. Platform logs document the exercised execution paths. A central open systems challenge is to turn evidence identity and fork lineage into a dependence model for correlated and adaptively selected AI branches.
Authors: Bryce Little
Abstract: Length-penalized reinforcement learning can shorten chain-of-thought reasoning while hiding an influence that drives the model's answer. In our experiments, training with length penalties does not stop misleading hints from steering models, even though the models' chains of thought mention the hint much less often. A token-accuracy evaluation would count these runs as successful because they use fewer reasoning tokens with little accuracy loss; it would miss whether the remaining trace still shows what drove the answer. We train Qwen3-4B and Qwen3-14B variants with different target chain lengths, then evaluate them with biasing-hint interventions on held-out MMLU-Pro-R and four transfer benchmarks. Compression sharply cuts reasoning tokens, preserves most multiple-choice accuracy, and leaves hint influence near baseline. At the strongest target, lower-bound faithfulness falls to 63.1% of baseline for Qwen3-14B and 69.4% for Qwen3-4B; the raw rate at which a monitor catches hint use falls from 69% to 49% and from 60% to 48%. To separate length from content, we randomly delete sentences from uncompressed baseline chains until the remaining text matches the compressed length. Even after this length matching, compressed chains disclose the hint 7-35 percentage points less often than baseline chains that we shorten at random, for both Qwen3 sizes and all five evaluation distributions. Compression therefore does more than shorten reasoning, preferentially removing the cues a monitor needs to see what influenced the answer. Together, these results reveal a compression-monitorability frontier in which cheaper reasoning can preserve answers while making the influences behind them harder to detect.
Authors: Jiayi Tian, Shiao Liu, Yuting Xu, Jia Lu, Zihao Guan, Honglin Han, Di Yang, Minqi Gu, Yifei Qian, Tianlin Zhang, Yanqing Zhu, Zeqian Ye, Menglin Yang, Fei Wang, Xu Hu, Xiuxian Li, Wei Zhang, Shihui Su, Yiyan Ji, Jingbo Wang, Ziteng Feng, Jiaheng Liu, Zhaoxiang Zhang, Xiaolong Wu, Zixiao Tang, Zhining Gu, Yang Cai, Linbo Zheng, Jingjing Ma, Mingyang Yin, Zedong Chu, Wenbin Tang, Mu Xu
Abstract: Recent VLM and VLA systems have improved robotic perception and action prediction, yet long-horizon embodied agents still require a general runtime layer for reasoning, memory, tool use, verification, and cross-embodiment execution. We present ABot-AgentOS, a general robotic Agent Operating System that sits above low-level controllers and provides a deliberative agent layer for scene-conditioned planning, context-isolated skill execution, multi-stage verification, multi-modal memory, and edge-cloud collaboration. To evaluate such systems, we introduce EmbodiedWorldBench, an executable benchmark with 16 indoor, outdoor, and hybrid scenes, four difficulty levels, and over 200 tasks involving navigation, object search, NPC dialogue, dynamic events, and trace-grounded scoring. ABot-AgentOS further introduces Universal Multi-modal Graph Memory, a persistent source-grounded substrate that converts dialogue, visual observations, spatial context, temporal relations, and task traces into typed nodes and edges. A failure-driven self-evolution loop converts diagnosed memory failures into gated runtime evo-assets that are promoted only to later evaluation splits, preventing current-split ground-truth leakage while enabling continual improvement. On an initial EmbodiedWorldBench subset, ABot-AgentOS improves over a single-controller baseline in both task success and goal completion. Across memory benchmarks, ABot-AgentOS Static achieves 87.5 on LoCoMo, 59.9 on OpenEQA EM-EQA, 88.6 on Mem-Gallery, and 76.5 Acc@All on NExT-QA; self-evolution further improves LoCoMo to 88.7, OpenEQA to 60.4, and Mem-Gallery to 89.0. These results suggest that a general Agent OS layer can improve long-horizon embodied execution while providing persistent, auditable memory for continual interaction.
Authors: Christoph Benzm\"uller, Daniel Kirchner
Abstract: We extend, in Isabelle/HOL, the deep-and-shallow embedding methodology of our prior work from propositional to first-order modal logic (FML) with constant-domain Kripke semantics. Three embeddings of FML into classical higher-order logic (HOL) are provided side by side: a deep embedding, a heavyweight maximal-shallow embedding, and a lightweight minimal-shallow embedding. The minimal-shallow embedding is presented as an Isabelle/HOL locale, parametrised by an accessibility relation, a world-indexed interpretation, a universe of worlds, and a variable assignment; the locale form admits a global faithfulness theorem, stating that quantifying over all minimal-shallow interpretations recovers exactly deep validity. A central technical contribution is a mechanisation, for FML under constant-domain Kripke semantics, of the (countable) downward L\"owenheim-Skolem theorem, which underpins the automation of our faithfulness proof between the deep and minimal-shallow embeddings. Deploying it inside an extension of the minimal-shallow locale resolves the surjectivity problem that arises against an uncountable domain of individuals -- where the locale's variable assignment, having countable domain V = nat, cannot be surjective onto the domain -- and thereby yields faithfulness over the full domain. Since prior work treats only the propositional fragment, we develop here the substitution machinery (free/bound-variable predicates, the fresh-variable function, capture-avoiding substitution, alphabetic renaming, the substitutability predicate, the substitution lemma, and size-based induction principles) needed for the first-order quantifiers.
Authors: Wencheng Ye, Yi Bin, Yujuan Ding, Hongye Fang, Zheng Wang, Xing Xu, Jingkuan Song, Yun Zhang, Sirui Da, Heng Tao Shen
Abstract: Vision-language models increasingly succeed on multimodal reasoning benchmarks, yet their visual evidence often becomes unstable once it enters the language stack, weakening evidence-grounded reasoning. To understand this fragility, we examine the internal dynamics of VLMs through a mechanistic lens and uncover a stable three-stage redistribution of multimodal attention focus across depth: an early question-conditioned organization, a critical middle visual-dominant relay, and a late return to answer formation. We operationalize the middle phase as the Visual Relay Window (VRW), and show that its geometry varies with task demand, is causally tied to grounded generation, and distinguishes unsupported answers from stronger reasoning trajectories. Guided by this internal rhythm, we propose TRACE, a task-adaptive inference-time control framework with lightweight trained modules. It reshapes relay allocation during prefill and preserves assembled visual support after handoff during decoding. Across four open-weight VLM backbones and seven benchmarks, TRACE delivers large gains on grounding-sensitive settings, improving them by 4.33 points on average and by up to 6.6 points, while also improving reasoning-heavy tasks. These results show that explicitly controlling multimodal focus across depth offers a unified and effective mechanism for strengthening evidence-grounded multimodal reasoning.
Authors: Pedro Orvalho, Guillem Aleny\`a, Felip Many\`a
Abstract: Vision--Language Models (VLMs) have recently demonstrated promising performance on structured visual reasoning tasks, including grid-based puzzles. However, despite strong perceptual capabilities, these models lack explicit mechanisms for enforcing logical consistency and frequently generate assignments that violate underlying constraints. In this paper, we propose a neuro-symbolic approach that integrates formal constraint reasoning into the VLM solving process via a Maximum Satisfiability (MaxSAT) oracle. Rather than computing solutions directly, the symbolic component acts as a consistency validator and refinement engine. Candidate placements generated by the VLM are encoded as soft clauses in a partial MaxSAT formulation, while Sudoku constraints remain hard clauses. When inconsistencies arise, the MaxSAT solver identifies a largest mutually consistent subset of assignments, which is then translated into structured textual and visual feedback to guide subsequent refinements. We evaluate our approach on a Sudoku dataset across multiple open-source and closed-access VLMs. Results show that MaxSAT-based feedback improves logical consistency and increases the number of solved instances, particularly in full-board refinement mode. These findings demonstrate that symbolic optimisation can enhance the reliability of vision-language reasoning.
Authors: Shiyu Ying, Xuejie Cao, Yingfan Ma, Yuanhao Dong, Wenyu Chen, Bowen Song, Lin Zhu
Abstract: Payment integration is a demanding repository-level software task: agents must select a suitable product, implement coordinated client-server flows, verify payment outcomes, and preserve consistency between transaction and business states. We introduce Alipay-PIBench, a benchmark for evaluating coding agents on realistic Alipay payment integration. It contains nine product-specific projects and 18 task instances, each organized into Basic functional-completion and Advanced risk-aware hardening scenarios. Scenario-specific rubrics support deterministic static, unit, integration, and end-to-end checks, supplemented by LLM-assisted assessment for semantic requirements. We evaluate six coding-agent models and report rubric pass rate (RPR). Under the with-skill condition, mean RPR ranges from 68.58% to 91.37%. Access to the alipay-payment-integration skill improves mean RPR by 10.31 percentage points on average relative to the without-skill condition, with gains varying across models, products, and scenarios. Method-level results distinguish source-level completion, executable payment behavior, and payment-domain requirements. Alipay-PIBench provides a controlled setting for diagnosing model capability and evaluating structured guidance in payment integration.
Authors: Haoxuan Li, Tianci Gao, Jianhe Li, Yang Fan, Runze Shi, Weiran Wang, Tianxiang Zhao, Zezhao Wu, Xiaoyang Jiang, Qihui Zhang, Jia Li, Xiao Xiao, Kai Du, Xiaoxuan Jia, Chao Xie, Lu Mi
Abstract: Understanding the brain increasingly depends on integrating evidence across scales, modalities, and disciplines. Addressing a single research question therefore requires a coordinated sequence of operations, from surveying prior work to executing analyses and interpreting results in light of domain knowledge. AI agents promise to accelerate this process, but current agents lack domain expertise in brain science, may fabricate claims, drift during multi-step reasoning, and offer few defined points for expert intervention. These failures are especially costly in brain science, where conclusions feed into downstream scientific claims and depend on laboratory-specific expertise and careful human judgment. We present \textbf{BrainPilot} a \textbf{fully open-source} multi-agent system that accelerates brain science research with traceable logs and agent-verified results. A principal investigator (PI) agent coordinates specialist agents grounded in curated domain knowledge: a unified brain science knowledge base containing 7{,}233 indexed items and a skill library of 72 reusable methodology units across seven research domains. Every major step is recorded in the Graph of Trace, an auditable record that links subgoals, tool use, evidence, and claims and allows researchers to follow and inspect the workflow. An Auditor agent further integrates fabrication checking into the workflow. For evaluation, we run three brain science tasks from Agents' Last Exam, introduce our own benchmark, \textbf{BrainPilotBench-v0}, and present additional end-to-end case studies. Across these evaluations, BrainPilot with an open-source backbone model attains performance comparable to state-of-the-art agent framework with less costs.
Authors: Vladimir Fedosov, Aleksandr Sazhin, Artemiy Grinenko, Frank Woernle
Abstract: Parameter-efficient fine-tuning reduces model and optimizer memory, but dense attention still makes long training sequences expensive. We combine Hierarchical Global Attention (HGA) with segment-wise backpropagation and tiered KV storage. Only the active segment remains differentiable in VRAM; older KV is detached into RAM or NVMe, and HGA loads a bounded set of exact historical tokens for each query block. On Qwen3-8B with 4-bit QLoRA and PG19, dense training on a 16 GB Quadro RTX 5000 fits 2,048 tokens but fails at 4,096, whereas HGA reaches 16,384 tokens with 15.28 GB peak VRAM. Under evaluation the same adapter runs through 131,072 tokens on this card; VRAM is not constant but grows gently with the resident chunk summaries, so RAM and NVMe capacity set the practical limit beyond these lengths. At the shared 2K training length, HGA-trained and dense-trained adapters obtain 2.7405 and 2.7383 nat under the same dense-attention readout, while the stock model obtains 2.9541. At this boundary HGA training is already marginally faster (217.75 vs. 207.02 tokens/s), and the HGA-to-dense throughput ratio improves from 1K to 2K; because HGA keeps the attended historical set per token approximately constant while dense work per token grows, we expect this lead to widen as context grows. Dense attention is used for the main quality and retrieval comparisons so that they measure the learned weights and remain compatible with standard generation frameworks. HGA can also be used for retrieval and generation; an optimized production-grade serving implementation is under development.
Authors: Han Jiang, Sunbeom Kwon, Jinwen Luo, Ziang Xiao, Susu Zhang
Abstract: AI benchmarks increasingly leverage item-level statistical models, particularly item response theory (IRT), to estimate model capabilities, rank systems, select informative examples, and diagnose benchmark quality. However, AI benchmark data often departs from the data regime of human testing, for which standard IRT estimation tools were originally developed: benchmarks typically involve fewer evaluated models, far more items, and capability distributions that may be skewed, clustered, or multimodal. We examine how these regime mismatches challenge the reliability of IRT modeling for AI evaluation. Using item parameters and capability distributions derived from six widely used LLM benchmarks, we simulate response matrices under three common IRT models and compare four estimation tools used in recent benchmark studies: marginal maximum likelihood, Markov chain Monte Carlo, variational inference, and a neural pseudo-Siamese estimator. Across 18,000 simulation conditions, we systematically evaluate computational feasibility, scalability, and the reliability of IRT inferences about model rankings, predicted performance, and item characteristics. Results show that classical estimators can become infeasible in large benchmark settings, whereas scalable estimators can produce unreliable item-level and ranking inferences with small or non-normally distributed model sets. This study identifies when latent trait models reliably support or risk distorting AI benchmarking claims, and what sample sizes and diagnostics are needed for trustworthy use.
Authors: Mohammad Eslami, Solale Tabarestani, Saber Kazeminasab, Ehsan Adeli, Glyn Elwyn, Tobias Elze, Mengyu Wang, Nazlee Zebardast, Lucia Sobrin, Nassir Navab, Daniel Shu Wei Ting, Malek Adjouadi
Abstract: Explainable Artificial Intelligence (XAI) is essential for trustworthy AI in healthcare, yet many existing methods rely on technical explanations that are difficult for clinicians and patients to interpret. We introduce Visualized Learning for Machine Learning (VL4ML), a human-centered explainability framework that communicates model predictions and uncertainty through intuitive visual representations rather than numerical or post-hoc explanations. By encoding diagnostic information in colors, patterns, and spatial structures, VL4ML enables users to interpret predictions without requiring knowledge of model internals or statistical expertise. We demonstrate the framework across multiple clinical tasks, including classification, regression, longitudinal prediction, and multimodal analysis. Its effectiveness was evaluated through a human-centered study involving 158 participants (39.2% clinical professionals) and an expert interpretability assessment. More than 79% of participants positively rated the visual explanations across evaluation dimensions, 84.0% found them more memorable than numeric outputs, and 76.9% reported faster decision-making. Over 82% successfully perceived uncertainty embedded in the visual representations without prior statistical training. No significant differences were observed between clinicians and non-clinicians or between male and female participants, indicating broad accessibility. These results suggest that VL4ML complements existing XAI and uncertainty quantification methods by providing intuitive, universally interpretable visual explanations that support transparent and trustworthy clinical decision-making.
Authors: Vinoth Nandakumar, Arush Tagade, Tongliang Liu
Abstract: Over the past decade deep learning has revolutionized the field of computer vision, with convolutional neural network models proving to be very effective for image classification benchmarks. However, a fundamental theoretical questions remain answered: why can they solve discrete image classification tasks that involve feature extraction? We address this question in this paper by introducing a novel mathematical model for image classification, based on feature extraction, that can be used to generate images resembling real-world datasets. We show that convolutional neural network classifiers can solve these image classification tasks with zero error. In our proof, we construct piecewise linear functions that detect the presence of features, and show that they can be realized by a convolutional network.
Authors: Haozheng Luo, Jiahao Yu, Wenxin Zhang, Jialong Li, Chenghao Qiu, Yimin Wang, Eric Hanchen Jiang, Jerry Yao-Chieh Hu, Yan Chen, Binghui Wang, Xinyu Xing, Han Liu
Abstract: We introduce a training-free safety enhancement method for aligning large language models (LLMs) without the need for supervised fine-tuning or reinforcement learning from human feedback. Our main idea is to provide a robust plug-and-play approach to prevent shadow alignment when models are adapted to downstream tasks. Specifically, we leverage knowledge distillation to extract alignment signals from well-aligned LLMs and inject them into shadow-aligned models via model fusion, enabling plug-and-play alignment correction. In our methodology, we employ delta debugging to identify the critical components of knowledge necessary for effective distillation. On the harmful question dataset, our method significantly enhances the average defense success rate by approximately 14.42%, reaching as high as 51.39% across 17 influenced LLMs, without compromising performance. Our code is available at https://github.com/NWULIST/DAPA.
Authors: Thomas Chen
Abstract: We derive explicit equations governing the cumulative biases and weights in Deep Learning with ReLU activation function, based on gradient descent for the Euclidean loss in the input layer, and under the assumption that the weights are, in a precise sense, adapted to the coordinate system distinguished by the activations. We show that gradient descent corresponds to a dynamical process in the input layer, whereby clusters of data are progressively reduced in complexity ("truncated") at an exponential rate that increases with the number of data points that have already been truncated. We provide a detailed discussion of several types of solutions to the gradient flow equations. A main motivation for this work is to shed light on the interpretability question in supervised learning.
Authors: Yurui Li, Yuxuan Chen, Xiaoli Yang, Shijian Li, Gang Pan
Abstract: The critical role of division of labor (DOL) in enhancing cooperation is well-recognized in real-world applications. Consequently, many cooperative multi-agent reinforcement learning (MARL) methods have incorporated DOL mechanisms to improve cooperation among agents. However, the lack of benchmark tasks specifically designed to evaluate and promote DOL and cooperation has limited the effective development and deployment of such mechanisms in cooperative MARL. This gap between current cooperative MARL methods and practical applications underscores the need for evaluation tasks that explicitly require DOL and cooperation. To address this gap, we propose the Composite Tasks Challenge (CTC), a suite of tasks explicitly designed to require both DOL and cooperation for successful task completion. The CTC tasks are constructed based on two core design principles: 1) DOL is a necessary condition for task success; 2) Failure in any atomic subtask results in failure of the overall task. The first principle emphasizes the necessity of DOL, while the second enforces the importance of cooperation, making both components essential for success in CTC tasks. We evaluate nine representative cooperative MARL methods on the proposed CTC tasks. Experimental results show that all methods consistently achieve zero test winning rates across all CTC tasks, highlighting the challenge of CTC tasks and the limitations of current methods. To facilitate future research, we also introduce a guiding solution that achieves non-zero test winning rates on all tasks, thereby demonstrating the solvability of the CTC tasks. However, the performance of this guiding solution remains suboptimal, further underscoring the value of CTC tasks as a challenging and meaningful testbed for advancing cooperative MARL research.
Authors: Haonan Yu, Junhao Liu, Xin Zhang
Abstract: Anchors is a popular local model-agnostic explanation technique whose applicability is limited by its computational inefficiency. To address this limitation, we propose a memorization-based framework that accelerates Anchors while preserving explanation fidelity and interpretability. Our approach leverages the iterative nature of Anchors' algorithm which gradually refines an explanation until it is precise enough for a given input by storing and reusing intermediate results obtained during prior explanations. Specifically, we maintain a memory of low-precision, high-coverage rules and introduce a rule transformation framework to adapt them to new inputs: the horizontal transformation adapts a pre-trained explanation to the current input by replacing features, and the vertical transformation refines the general explanation until it is precise enough for the input. We evaluate our method across tabular, text, and image datasets, demonstrating that it significantly reduces explanation generation time while maintaining fidelity and interpretability, thereby enabling the practical adoption of Anchors in time-sensitive applications.
Authors: Yuni Lai, Yulin Zhu, Yixuan Sun, Yulun Wu, Bin Xiao, Gaolei Li, Jianhua Li, Qi Xie, Kai Zhou
Abstract: Despite advancements in Graph Neural Networks (GNNs), adaptive attacks continue to challenge their robustness. Certified robustness via randomized smoothing offers provable guarantees but suffers from a severe accuracy-robustness trade-off, limiting its practical use. To bridge this gap, we introduce AuditVotes, the first framework that simultaneously achieves high clean accuracy and strong certified robustness. AuditVotes seamlessly integrates two novel components into the randomized smoothing pipeline: (1) graph rewiring augmentation, which denoises randomized graphs to recover data quality, and (2) conditional smoothing, which filters low-confidence votes to ensure prediction consistency. We establish a novel theoretical result, proving that certified robustness is preserved under arbitrary filtering functions. Designed for inductive learning, our framework generalizes to unseen nodes and applies broadly to other smoothing schemes, including de-randomized smoothing for graphs and Gaussian smoothing for images. Extensive experiments show AuditVotes delivers substantial gains: on Cora-ML under 20-edge attacks, it improves clean accuracy by 437.1% and certified accuracy by 409.3%, while maintaining comparable runtime to vanilla smoothing. As a widely applicable and efficient plug-in, AuditVotes offers higher accuracy and stronger guarantees, enabling the practical and certifiably robust GNNs in security-sensitive domains.
Authors: Ethan Dickey, Andres Bejarano, Rhianna Kuperus, B\'arbara Fagundes
Abstract: Background and Context. Generative AI (GenAI) tools are increasingly used in programming courses, but we have limited evidence about how brief instruction can foster responsible, learning-oriented use. Objectives. We evaluate "AI-Lab", a scaffolded GenAI literacy intervention, asking how students' self-reported GenAI usage and their openness and comfort using GenAI for conceptual, debugging, and homework tasks change after participation. Methods. Across two semesters in three CS courses and one first-year engineering course at a U.S. university, we deployed the "AI-Lab" (pre-lab orientation, in-class critique of GenAI outputs, and a required homework reflection), collecting paired pre/post surveys (Perception N=831; Usage N=826) and six post-intervention focus groups; primary inferential analyses used the three CS courses (N=778 and 773, respectively). We analyzed survey shifts with paired non-parametric tests and focus groups via thematic analysis. Findings. Openness increased for conceptual questions and homework help, and comfort increased for conceptual, debugging, and homework scenarios; self-reported frequency of GenAI use for homework and projects remained stable, while self-reported use for debugging increased. Focus group participants described adopting more iterative prompting strategies, becoming more skeptical of correctness, and articulating clearer boundaries around integrity and dependence. Implications. A short, structured intervention can shift students' reported comfort with and willingness to use GenAI and influence the strategies they describe for engaging with it without increasing overall self-reported use on graded work. These results motivate future work triangulating surveys with behavioral traces and learning measures.
Authors: Jiaheng Hu, Peter Stone, Roberto Mart\'in-Mart\'in
Abstract: Building capable household and industrial robots requires mastering the control of versatile, high-degree-of-freedom (DoF) systems such as mobile manipulators. While reinforcement learning (RL) holds promise for autonomously acquiring robot control policies, scaling it to high-DoF embodiments remains challenging. Direct RL in the real world demands both safe exploration and high sample efficiency, which are difficult to achieve in practice. Sim-to-real RL, on the other hand, is often brittle due to the reality gap. This paper introduces SLAC, a method that renders real-world RL feasible for complex embodiments by leveraging a low-fidelity simulator to pretrain a task-agnostic latent action space. SLAC trains this latent action space via a customized unsupervised skill discovery method designed to promote temporal abstraction, disentanglement, and safety, thereby facilitating efficient downstream learning. Once a latent action space is learned, SLAC uses it as the action interface for a novel off-policy RL algorithm to autonomously learn downstream tasks through real-world interactions. We evaluate SLAC against existing methods on a suite of bimanual mobile manipulation tasks, where it achieves state-of-the-art performance. Notably, SLAC learns contact-rich whole-body tasks in under an hour of real-world interactions, without relying on any demonstrations or hand-crafted behavior priors. More information and robot videos at robo-rl.github.io
Authors: Andris Ambainis, Joao F. Doriguello, Debbie Lim
Abstract: We propose novel classical and quantum online algorithms for learning finite- and infinite-horizon Markov Decision Processes (MDPs). Our algorithms are based on a hybrid online-offline reinforcement learning model wherein the agent can, from time to time, freely interact with the environment in a generative sampling fashion, i.e., by having access to a "simulator". By employing known classical and new quantum algorithms for approximating optimal policies under a generative model within our learning algorithms, we show that it is possible to avoid several paradigms from RL like "optimism in the face of uncertainty" and "posterior sampling" and instead compute and use optimal policies directly, which yields better regret bounds compared to previous works. Our quantum algorithms obtain regret bounds which only a $\operatorname{poly}\log{T}$ dependence on the number of time steps $T$, thus breaking the $O(\sqrt{T})$ classical barrier. Our infinite-horizon discounted regret bound is brand new, while in the finite- and infinite-horizon undiscounted settings, our results match the time dependence of some prior quantum works, but with improved dependence on other parameters like state space size $S$ and action space size $A$.
Authors: Belman Jahir Rodriguez, Sergio F. Chevtchenko, Marcelo Herrera Martinez, Yeshwanth Bethi, Saeed Afshar
Abstract: We introduce a U-net model for 360{\deg} acoustic source localization formulated as a spherical semantic segmentation task. Rather than regressing discrete direction-of-arrival (DoA) angles, our model segments beamformed audio maps (azimuth & elevation) into regions of active sound presence. Using delay-and-sum (DAS) beamforming on a custom 24-microphone array, we generate signals aligned with drone GPS telemetry to create binary supervision masks. A modified U-Net, trained on frequency-domain representations of these maps, learns to identify spatially distributed source regions while addressing class imbalance via the Tversky loss. Because the network operates on beamformed energy maps, the approach is inherently array-independent and can adapt to different microphone configurations and can be transferred to different microphone configurations with minimal adaptation. The segmentation outputs are post-processed by computing centroids over activated regions, enabling robust DoA estimates. Our dataset includes real-world open-field recordings of a DJI Air 3 drone, synchronized with 360{\deg} video and flight logs across multiple dates and locations. Experimental results show that U-net generalizes across environments, providing improved angular precision, offering a new paradigm for dense spatial audio understanding beyond traditional Sound Source Localization (SSL). We additionally validate the same beamforming-plus-segmentation formulation on the DCASE 2019 TAU Spatial Sound Events benchmark, showing that the approach generalizes beyond drone acoustics to multiclass Sound Event Localization and Detection (SELD) scenarios.
Authors: Laura Hellwege, Johann Christopher Engster, Moritz Schaar, Thorsten M. Buzug, Maik Stille
Abstract: Assume you encounter an inverse problem that shall be solved for a large number of data, but no ground-truth data is available. To emulate this, in this study we assume it is unknown how to solve the imaging problem of Computed Tomography. We introduce an unsupervised deep learning framework that leverages the inherent similarities between iterative reconstruction, Deep Image Prior (DIP), and unrolled optimization schemes. Our specific contribution is a training framework for amortized reconstruction: After training on a dataset without any image-domain ground truth, reconstruction of an unseen scan reduces to a single network forward pass. We demonstrate the feasibility of reconstructing images from measurement data by pure network inference, without additional gradient steps for unseen samples. Our method is evaluated on the two-dimensional 2DeteCT dataset. Within a controlled, geometry-matched benchmark, our reconstructions are competitive with, or better than, filtered back-projection, maximum-likelihood reconstruction, and a supervised network of identical architecture. Compared to a per-image DIP baseline, our method reaches similar quality while replacing the costly per-instance optimization with a single forward pass, yielding a speed-up of about four orders of magnitude. This makes it a promising candidate for time-critical imaging applications. Future work will address multi-dataset adaptability, counter-measures against over-smoothing, advanced uncertainty quantification, and further medical-imaging inverse problems.
Authors: Wenpeng Xing, Bohan Yang, Mohan Li, Chunqiang Hu, Haitao Xu, Ningyu Zhang, Bo Lin, Meng Han
Abstract: Safety-aligned large language models can still be manipulated through white-box interventions that modify their internal representations. We introduce Latent Fusion Jailbreak (LFJ), which works by pairing a harmful query with a structurally similar but benign counterpart, then interpolating their hidden states at carefully selected layers and token positions. Refusal-loss gradients determine exactly where to intervene, and we optimise layer-wise mixing coefficients using token-normalised compliance and refusal-suppression objectives. The edited prompt states propagate sequentially through the remaining transformer blocks. Across four safety benchmarks and five open-weight target models, LFJ reaches a macro-averaged attack success rate (ASR) of 94.13% under the white-box protocol we describe. Because LFJ directly accesses internal states, comparisons with prompt-only attacks serve as a descriptive reference rather than a matched evaluation. Dropping rejection sampling lowers ASR to 86.72%, whereas replacing the structured harmful-benign pairing with random pairing causes it to fall to 27.45%. We also design an LFJ-specific latent adversarial training procedure that, when the attack is re-optimised against the defended model, reduces ASR from 94.13% to 12.37%. This defence evaluation does not cover transfer to other attack types or preservation of benign utility.
Authors: Soumia Zohra El Mestari, Maciej Krzysztof Zuziak, Gabriele Lenzini
Abstract: Federated Learning (FL) enables collaborative model training among clients without centralising data, making it a widely adopted privacy-enhancing technology (PET). Despite its privacy benefits, FL remains vulnerable to orchestrator-driven privacy attacks. In this paper, we study an underexplored threat in which a dishonest orchestrator intentionally manipulates the aggregation process to induce targeted overfitting in local models of specific clients. Although prior work focuses on reducing information leakage during training, we emphasise early client-side detection of targeted overfitting, allowing clients to disengage before significant harm occurs. To this end, we propose three detection techniques -- label flipping, backdoor trigger injection, and model fingerprinting -- which enable clients to verify the integrity of the global aggregation. We evaluated our methods across multiple datasets and attack scenarios. In single-client attacks, all three methods detect orchestrator-induced overfitting within 1-2 training rounds with F1 scores up to 0.7. Scalability experiments further show that detection effectiveness is influenced by cohort composition and method parameters. These results demonstrate that client-side integrity testing can provide early, effective, and scalable detection, supporting safer deployment of FL systems.
Authors: Jorge Mendez-Mendez
Abstract: While we know that large language models (LLMs) can solve some planning problems, we do not understand the extent of these capabilities for robotics. One promising direction is to integrate the semantic knowledge of LLMs with the formal reasoning of task and motion planning (TAMP). However, designing such systems is complicated by the myriad of choices for how to integrate LLMs within TAMP. We develop 16 algorithms that use LLMs to substitute key TAMP components. Our zero-shot experiments across 13750 evaluations and three domains reveal that LLM-based planners exhibit lower success rates and higher planning times than engineered systems. Providing geometric details increases the number of task-planning errors compared to pure PDDL descriptions, and (faster) direct LLM variants outperform (slower) reasoning variants in most cases. Code and results are available at https://github.com/jorge-a-mendez/llm-pddlstream.
Authors: Xiao Yang, Xuejiao Zhao, Zhiqi Shen
Abstract: Graph neural networks (GNNs) have achieved remarkable success in node classification. Building on this progress, heterogeneous graph neural networks (HGNNs) integrate relation types and node and edge semantics to leverage heterogeneous information. Causal analysis for HGNNs is advancing rapidly, aiming to separate genuine causal effects from spurious correlations. However, whether HGNNs are intrinsically effective for node classification remains underexamined, and most studies implicitly assume rather than establish this effectiveness. In this work, we examine HGNNs for node classification from two perspectives: model architecture and heterogeneous information. We conduct a systematic reproduction across 21 datasets and 20 baselines, complemented by comprehensive hyperparameter retuning. To further disentangle the source of performance gains, we develop a causal mediation analysis framework that treats the introduction of heterogeneous relation information as the treatment, candidate structural properties as mediators, and node classification performance as the outcome. This framework first screens candidate mediators according to their treatment-induced changes and their associations with performance improvement, and then decomposes the total effect into mediated and direct effects. Our results lead to two conclusions. First, model architecture and complexity have no causal effect on node classification performance. Second, heterogeneous information exerts a positive causal effect primarily through increasing homophily and local-global distribution discrepancy, which makes node classes more distinguishable. The implementation is publicly available at https://github.com/YXNTU/CausalHGNN.
Authors: Bianca Raimondi, Daniela Dalbagno, Maurizio Gabbrielli
Abstract: Large language models (LLMs) have been shown to internalize human-like biases during finetuning, yet the mechanisms by which these biases manifest remain unclear. In this work, we investigated whether the well-known Knobe effect, a moral bias in intentionality judgements, emerges in finetuned LLMs and whether it can be traced back to specific components of the model. We conducted a Layer-Patching analysis across 3 open-weights LLMs and demonstrated that the bias is not only learned during finetuning but also localized in a specific set of layers. Surprisingly, we found that patching activations from the corresponding pretrained model into just a few critical layers is sufficient to eliminate the effect. Our findings offer new evidence that social biases in LLMs can be interpreted, localized, and mitigated through targeted interventions, without the need for model retraining.
Authors: Weixian Qian, Tianyi Yang, Sebastian Schroder, Yao Deng, Jiaohong Yao, Xiao Cheng, Richard Han, Xi Zheng
Abstract: Reliable assessment of safe landing sites in unstructured environments is essential for deploying Unmanned Aerial Vehicles (UAVs) in real-world applications such as delivery, inspection, and surveillance. Existing learning-based approaches often degrade under covariate shift and offer limited transparency, making their decisions difficult to interpret and validate on resource-constrained platforms. We present NeuroSymLand, a neuro-symbolic framework for marker-free UAV landing site safety assessment that explicitly separates perception-driven world modeling from logic-based safety reasoning. A lightweight segmentation model incrementally constructs a probabilistic semantic scene graph encoding objects, attributes, and spatial relations. Symbolic safety rules, synthesized offline via large language models with human-in-the-loop refinement, are executed directly over this world model at runtime to perform white-box reasoning, producing ranked landing candidates with human-readable explanations of the underlying safety constraints. Across 72 simulated and hardware-in-the-loop landing scenarios, NeuroSymLand achieves 61 successful assessments, outperforming four competitive baselines, which achieve between 37 and 57 successes. Qualitative analysis highlights its superior interpretability and transparent reasoning, while deployment incurs negligible edge overhead. Our results suggest that combining explicit world modeling with symbolic reasoning can support accurate, interpretable, and edge-deployable safety assessment in mobile systems, as demonstrated through UAV landing site assessment.
Authors: Vaibhav Singh, Oleksiy Ostapenko, Pierre-Andr\'e No\"el, Eugene Belilovsky, Torsten Scholak
Abstract: Diffusion language models (DLMs) have emerged as a promising alternative to autoregressive (AR) generation, yet their reliance on Transformer backbones limits inference efficiency due to quadratic attention or KV-cache overhead. We introduce DiffuMamba, a masked diffusion language model built on a bidirectional Mamba backbone that combines the diffusion objective with linear-time sequence modeling, and DiffuMamba-H, a hybrid variant with interleaved attention. Across scales up to 1.3B parameters, our models match Transformer-based diffusion in downstream performance while achieving up to 8.2x and 4.3x higher inference throughput, respectively, on long sequences. We further present a systematic analysis of inference efficiency across modern DLM variants combining asymptotic complexity with empirical measurements. Notably, cache-efficient block diffusion with Mamba mixers emerges as the only strategy that scales linearly with sequence length and achieves the strongest performance across all baselines, suggesting a promising direction for future diffusion-based generation systems.
Authors: Jiazhao Shi, Pan Pan, Haotian Shi
Abstract: This article trained a network for perceiving three-dimensional motion information of binocular vision target, which can provide real-time three-dimensional coordinate, velocity, and acceleration, and has a basic spatiotemporal perception capability. Understood the ability of neural networks to fit nonlinear problems from the perspective of PID. Considered a single-layer neural network as using a second-order difference equation and a nonlinearity to describe a local problem. Multilayer networks gradually transform the raw representation to the desired representation through multiple such combinations. Analysed some reference principles for designing neural networks. Designed a relatively small PID convolutional neural network, with a total of 17 layers and 413 thousand parameters. Implemented a simple but practical feature reuse method by concatenation and pooling. The network was trained and tested using the simulated randomly moving ball datasets, and the experimental results showed that the prediction accuracy was close to the upper limit that the input image resolution can represent. Analysed the experimental results and errors, as well as the existing shortcomings and possible directions for improvement. Finally, discussed the advantages of high-dimensional convolution in improving computational efficiency and feature space utilization. As well as the potential advantages of using PID information to implement memory and attention mechanisms.
Authors: Irina Tezaur, Eric Parish, Anthony Gruber, Ian Moore, Christopher Wentland, Alejandro Mota
Abstract: This paper presents a novel hybrid approach for coupling subdomain-local non-intrusive Operator Inference (OpInf) reduced order models (ROMs) with each other and with subdomain-local high-fidelity full order models (FOMs) with using the overlapping Schwarz alternating method (O-SAM). The proposed methodology addresses significant challenges in multiscale modeling and simulation, particularly the long runtime and complex mesh generation requirements associated with traditional high-fidelity simulations. By leveraging the flexibility of O-SAM, we enable the seamless integration of disparate models, meshes, and time integration schemes, enhancing computational efficiency while maintaining high accuracy. Our approach is demonstrated through a series of numerical experiments on complex three-dimensional (3D) solid dynamics problems, showcasing speedups of up to 106x compared to conventional FOM-FOM couplings. This work paves the way for more efficient simulation workflows in engineering applications, with potential extensions to a wide range of partial differential equations.
Authors: Ciro Benito Raggio, Lucia Migliorelli, Nils Skupien, Mathias Krohmer Zabaleta, Oliver Blanck, Francesco Cicone, Giuseppe Lucio Cascini, Paolo Zaffino, Maria Francesca Spadea
Abstract: Federated Learning (FL) holds the potential to advance equality in health by enabling diverse institutions to collaboratively train deep learning (DL) models, even with limited data. However, the significant resource requirements of FL often exclude centres with limited computational infrastructure, further widening existing healthcare disparities. To address this issue, we propose a Green AI-oriented adaptive layer-freezing strategy designed to reduce energy consumption and computational load while maintaining model performance. We tested our approach using different federated architectures for Magnetic Resonance Imaging (MRI)-to-Computed Tomography (CT) conversion. The proposed adaptive strategy optimises the federated training by selectively freezing the encoder weights based on the monitored relative difference of the encoder weights from round to round. A patience-based mechanism ensures that freezing only occurs when updates remain consistently minimal. The energy consumption and CO2eq emissions of the federation were tracked using the CodeCarbon library. Compared to equivalent non-frozen counterparts, our approach reduced training time, total energy consumption and CO2eq emissions by up to 23%. At the same time, the MRI-to-CT conversion performance was maintained, with only small variations in the Mean Absolute Error (MAE). Notably, for three out of the five evaluated architectures, no statistically significant differences were observed, while two architectures exhibited statistically significant improvements. Our work aligns with a research paradigm that promotes DL-based frameworks meeting clinical requirements while ensuring climatic, social, and economic sustainability. It lays the groundwork for novel FL evaluation frameworks, advancing privacy, equity and, more broadly, justice in AI-driven healthcare.
Authors: Zhiyu Xu, Jia Liu, Yixin Wang, Yuqi Gu
Abstract: The proliferation of Large Language Models (LLMs) necessitates valid evaluation methods to provide guidance for both downstream applications and actionable future improvements. The Item Response Theory (IRT) model with Computerized Adaptive Testing has recently emerged as a promising framework for evaluating LLMs via their response accuracy. Beyond simple response accuracy, LLMs' chain of thought (CoT) lengths serve as a vital indicator of their reasoning ability. To leverage the CoT length information to assist LLM evaluation, we propose the \textbf{La}tency-\textbf{R}esponse \textbf{T}heory (LaRT) model, which jointly models both the response accuracy and CoT length by introducing a key correlation parameter between the latent ability and the latent speed. We derive an efficient stochastic approximation Expectation-Maximization algorithm for parameter estimation. We establish rigorous identifiability results for the latent ability and latent speed parameters to ensure the statistical validity of their estimation. Through both theoretical asymptotic analyses and simulation studies, we demonstrate LaRT's advantages over IRT in terms of superior estimation accuracy and shorter confidence intervals for latent trait estimation. To evaluate LaRT in real data, we collect responses from diverse LLMs on popular benchmark datasets. We find that LaRT yields different LLM rankings than IRT and outperforms IRT across multiple key evaluation metrics including predictive power, item efficiency, ranking validity, and LLM evaluation efficiency. Code and data are available at https://github.com/Toby-X/Latency-Response-Theory-Model
URLs: https://github.com/Toby-X/Latency-Response-Theory-Model
Authors: Zhiyan Hou, Haiyun Guo, Haokai Ma, Yandu Sun, Yonghui Yang, Jinqiao Wang
Abstract: Continual instruction tuning (CIT) requires multimodal large language models (MLLMs) to adapt to a stream of tasks without forgetting prior capabilities. A common strategy is to isolate updates by routing inputs to different LoRA experts. However, existing LoRA-based Mixture-of-Experts (MoE) methods often jointly update the router and experts in an indiscriminate way, causing the router's preferences to co-drift with experts' adaptation pathways and gradually deviate from early-stage input--expert specialization. We term this as Misaligned Co-drift, which blurs expert responsibilities and exacerbates forgetting. To address this, we introduce the pathway activation subspace (PASs), a LoRA-induced subspace that reflects which low-rank pathway directions an input activates in each expert, providing a capability-aligned coordinate system for routing and preservation. Based on PASs, we propose a fixed-capacity PASs-based MoE--LoRA method with two components: PAS-guided Reweighting, which calibrates routing using each expert's pathway activation signals, and PAS-aware Rank Stabilization, which selectively stabilizes rank directions important to previous tasks. Experiments on a CIT benchmark show that our approach consistently outperforms a range of conventional continual learning baselines and MoE--LoRA variants in both accuracy and resistance to forgetting, without increasing model parameters. Our code is publicly available at https://github.com/yueluoshuangtian/PASs-MoE.
Authors: Charles Westphal, Keivan Navaie, Fernando E. Rosas
Abstract: Fine-tuned LLMs can covertly encode prompt secrets into outputs via steganographic channels. Prior work demonstrated this threat but relied on trivially recoverable encodings. We formalize payload recoverability via classifier accuracy and show previous schemes achieve 100\% recoverability. In response, we introduce low-recoverability steganography, replacing arbitrary mappings with embedding-space-derived ones. For Llama-8B (LoRA) and Ministral-8B (LoRA) trained on TrojanStego prompts, exact secret recovery rises from 17$\rightarrow$30\% (+78\%) and 24$\rightarrow$43\% (+80\%) respectively, while on Llama-70B (LoRA) trained on Wiki prompts, it climbs from 9$\rightarrow$19\% (+123\%), all while reducing payload recoverability. We then discuss detection. We argue that detecting fine-tuning-based steganographic attacks requires approaches beyond traditional steganalysis. Standard approaches measure distributional shift, which is an expected side-effect of fine-tuning. Instead, we propose a mechanistic interpretability approach: linear probes trained on later-layer activations detect the secret with up to 33\% higher accuracy in fine-tuned models compared to base models, even for low-recoverability schemes. This suggests that malicious fine-tuning leaves actionable internal signatures amenable to interpretability-based defenses.
Authors: Chenyi Ji, Kian P. Abdolazizi, Hagen Holthusen, Christian J. Cyron, Kevin Linka
Abstract: A key problem of solid mechanics is the identification of the constitutive law of a material, that is, the relation between strain history and stress. Machine learning has lead to considerable advances in this field lately. Here we introduce inelastic Constitutive Kolmogorov-Arnold Networks (iCKANs). This novel artificial neural network architecture can discover in an automated manner symbolic constitutive laws describing both the elastic and inelastic behavior of materials. That is, it can translate data from material testing into corresponding elastic and inelastic potential functions in closed mathematical form. We demonstrate the advantages of iCKANs using both synthetic data and experimental data of the viscoelastic polymer materials VHB 4910 and VHB 4905. The results demonstrate that iCKANs accurately capture complex viscoelastic behavior while preserving physical interpretability. It is a particular strength of iCKANs that they can process not only mechanical data but also arbitrary additional information available about a material (e.g., about temperature-dependent behavior). This makes iCKANs a powerful tool to discover in the future also how specific processing or service conditions affect the properties of materials.
Authors: Zhicheng Fang, Jingjie Zheng, Chenxu Fu, Wei Xu
Abstract: Jailbreak techniques for large language models (LLMs) evolve faster than benchmarks, making robustness estimates stale and difficult to compare across papers due to drift in datasets, harnesses, and judging protocols. We introduce JAILBREAK FOUNDRY (JBF), a system that addresses this gap via a multi-agent workflow to translate jailbreak papers into executable modules for immediate evaluation within a unified harness. JBF features three core components: (i) JBF-LIB for shared contracts and reusable utilities; (ii) JBF-FORGE for the multi-agent paper-to-module translation; and (iii) JBF-EVAL for standardizing evaluations. Across 30 reproduced attacks, JBF achieves high fidelity with a mean (reproduced-reported) attack success rate (ASR) deviation of +0.26 percentage points. By leveraging shared infrastructure, JBF reduces attack-specific implementation code by more than half relative to original repositories and achieves an 82.5% mean reused-code ratio. This system enables a standardized AdvBench evaluation of all 30 attacks across 10 victim models using a consistent GPT-4o judge. By automating both attack integration and standardized evaluation, JBF offers a scalable solution for creating living benchmarks that keep pace with the rapidly shifting security landscape.
Authors: Songming Zhang, Xue Zhang, Tong Zhang, Bojie Hu, Yufeng Chen, Jinan Xu
Abstract: Knowledge distillation (KD) is an essential technique to compress large language models (LLMs) into smaller ones. However, despite the distinct roles of the student model and the teacher model in KD, most existing frameworks still use a homogeneous training backend (e.g., FSDP and DeepSpeed) for both models, leading to suboptimal training efficiency. In this paper, we present a novel framework for LLM distillation, termed \textbf{KDFlow}, which features a decoupled architecture and employs SGLang for teacher inference. By bridging the training efficiency of FSDP2 and the inference efficiency of SGLang, KDFlow achieves full utilization of both advantages in a unified system. Moreover, instead of transferring full logits across different processes, our framework only transmits the teacher's hidden states using zero-copy data transfer and recomputes the logits on the student side, effectively balancing the communication cost and KD performance. Furthermore, our framework supports both off-policy and on-policy distillation and incorporates KD algorithms for cross-tokenizer KD through highly extensible and user-friendly APIs. Experiments show that KDFlow can achieve \textbf{1.44$\times$ to 6.36$\times$} speedup compared to current KD frameworks, enabling researchers to rapidly prototype and scale LLM distillation with minimal engineering overhead. Code is available at: https://github.com/songmzhang/KDFlow
Authors: Hao Zhang, Yves Tseng, Ding Zhao, H. Eric Tseng
Abstract: Cooperative object transport in unstructured environments remains challenging for assistive humanoids because strong, time-varying interaction forces can make tracking-centric whole-body control unreliable, especially in close-contact support tasks. This paper proposes a bio-inspired, interaction-oriented whole-body control (IO-WBC) that functions as an artificial cerebellum - an adaptive motor agent that translates upstream (skill-level) commands into stable, physically consistent whole-body behavior under contact. This work structurally separates upper-body interaction execution from lower-body support control, enabling the robot to maintain balance while shaping force exchange in a tightly coupled robot-object system. A trajectory-optimized reference generator (RG) provides a kinematic prior, while a reinforcement learning (RL) policy governs body responses under heavy-load interactions and disturbances. The policy is trained in simulation with randomized payload mass/inertia and external perturbations, and deployed via asymmetric teacher-student distillation so that the student relies only on proprioceptive histories at runtime. Extensive experiments demonstrate that IO-WBC maintains stable whole-body behavior and physical interaction even when precise velocity tracking becomes infeasible, enabling compliant object transport across a wide range of scenarios.
Authors: Mahmoud Ibrahim, Bart Elen, Chang Sun, Gokhan Ertaylan, Michel Dumontier
Abstract: Generative models are increasingly used to augment medical imaging datasets for fairer AI, yet a key assumption often goes unexamined: that generators produce equally high-quality images across demographic groups. Models trained on imbalanced data inherit these imbalances, degrading synthesis for rare subgroups and struggling with intersections absent from training: the imbalanced generator problem. Remedies such as loss reweighting operate at the optimization level and provide limited benefit when training signal is scarce or absent. We propose CompDiff, a hierarchical compositional diffusion framework that addresses this at the representation level. A dedicated Hierarchical Conditioner Network (HCN) decomposes demographic conditioning into single-attribute, pairwise, and composed representations, producing a demographic token concatenated with CLIP embeddings as cross-attention context. This structured factorization encourages parameter sharing across subgroups and supports compositional generalization to rare or unseen intersections. On chest X-rays (MIMIC-CXR) and fundus images (FairGenMed), CompDiff compares favorably against standard fine-tuning and FairDiffusion across image quality (FID 64.3 vs. 75.1), subgroup equity (ES-FID), and zero-shot intersectional generalization (up to 21% FID improvement on held-out intersections). Downstream classifiers trained on CompDiff data show improved AUROC and reduced demographic bias, suggesting that the architectural design of demographic conditioning is an important and underexplored factor in fair medical image generation. Code: https://github.com/mahmoudibrahim98/CompDiff.
Authors: Juan Gabriel Kostelec, Qinghai Guo
Abstract: Converting a pretrained Transformer into a more efficient hybrid model through distillation offers a promising approach to reducing inference costs. However, achieving high-quality generation in distilled models requires careful joint design of both the student architecture and the distillation process. Many prior distillation works evaluate downstream multiple-choice benchmarks by ranking candidate answers with log-likelihood rather than requiring autoregressive generation, which can obscure important differences in model quality. For example, on overlapping benchmarks, we show that a 7B distilled model that nearly matches its teacher to within 0.2 pp under log-likelihood scoring falls behind by 20.8 pp when it must generate answers autoregressively. We investigate this phenomenon with GenDistill, a multi-stage pipeline we designed for distilling a pretrained Transformer into an efficient Hybrid Kimi Delta Attention (Hybrid-KDA) student. Using it as a controlled testbed on Qwen3-0.6B, we systematically ablate six design axes (training objective, loss masking, training duration, dataset selection, parameter freezing, and architecture choice) and evaluate every choice under both log-likelihood and generation-based protocols. We find that log-likelihood-based evaluation consistently underestimates the gap between teacher and student, and can in some cases reverse the ranking of design choices, so conclusions drawn from perplexity-only evaluation may be misleading. Among the factors we study, dataset selection, completion-only masking, and freezing attention layers during post-training have the largest impact on generation quality. Our best distillation recipe, using a Hybrid-KDA model as the student, retains 86-90% of teacher accuracy on knowledge benchmarks while reducing KV cache memory by up to 75% and improving time-to-first-token by 2-4x at 128K-token contexts.
Authors: Byeolhee Kim, Min-Kyung Kim, Young-Hak Kim, Tae-Joon Jeon
Abstract: Retrieval-augmented generation (RAG) grounds large language models in external medical knowledge, yet standard retrievers frequently surface hard negatives that are semantically close to the query but describe clinically distinct conditions. While existing query-expansion methods improve query representation to mitigate ambiguity, they typically focus on enriching target-relevant semantics without an explicit mechanism to selectively suppress specific, clinically plausible hard negatives. This leaves the system prone to retrieving plausible mimics that overshadow the actual diagnosis, particularly when such mimics are dominant within the corpus. We propose Contrastive Hypothesis Retrieval (CHR), a framework inspired by the process of clinical differential diagnosis. CHR generates a target hypothesis $H^+$ for the likely correct answer and a mimic hypothesis $H^-$ for the most plausible incorrect alternative, then scores documents by promoting $H^+$-aligned evidence while penalizing $H^-$-aligned content. Across three medical QA benchmarks and three answer generators, CHR outperforms all five baselines in every configuration, with improvements of up to 10.4 percentage points over the next-best method. On the $n=587$ pooled cases where CHR answers correctly while embedded hypothetical-document query expansion does not, 85.2\% have no shared documents between the top-5 retrieval lists of CHR and of that baseline, consistent with substantive retrieval redirection rather than light re-ranking of the same candidates. By explicitly modeling what to avoid alongside what to find, CHR bridges clinical reasoning with retrieval mechanism design and offers a practical path to reducing hard-negative contamination in medical RAG systems.
Authors: Ruida Hu, Xinchen Wang, Chao Peng, Cuiyun Gao, David Lo
Abstract: The evolution of Large Language Models (LLMs) has catalyzed a paradigm shift towards intent-driven software development, where autonomous agents are expected to design and deliver complete, runnable software systems from scratch. However, existing benchmarks fail to adequately assess this 0-to-1 generation capability due to two fundamental limitations. First, they rely on predefined structural scaffolds, which reduces the task to mere file-filling. Second, they depend on rigid white-box unit testing, which forces generated code to conform to specific internal implementations rather than validating end-to-end user-centric behavior. To bridge this gap, we introduce CLI-Tool-Bench, a novel, structure-agnostic benchmark designed to evaluate the ground-up generation of Command-Line Interface (CLI) tools. Powered by an automated black-box differential testing framework, the benchmark comprises 94 high-quality, real-world repositories spanning diverse programming languages and complexity levels. For each task, agents are provided with an empty workspace, forcing them to autonomously handle repository planning and dependencies. We evaluate the generated software by executing it in isolated sandboxes. The system-level side effects and terminal outputs are then compared against human-written oracles using a rigorous multi-tiered equivalence metric. Extensive evaluation of seven state-of-the-art LLMs reveals that the top-tier models achieve a maximum overall success rate of only 43.8%, highlighting that 0-to-1 software generation remains a highly challenging frontier. Furthermore, we discover that agents exhibit a strong tendency to generate monolithic code structures, and that higher token consumption does not necessarily yield better task performance.
Authors: Alkesh Patel, Melis Ozyildirim, Ying-Chang Cheng, Ganesh Nagarajan
Abstract: Long video summarization presents significant challenges for multimodal large language models (MLLMs), particularly in maintaining temporal fidelity over extended durations and producing summaries that are both semantically and temporally grounded. We introduce LVSum, a human-annotated benchmark for evaluating long-form video summarization with fine-grained temporal alignment. LVSum comprises 72 diverse videos spanning 13 domains with an average duration of 16 minutes, each annotated with up to 10 human-generated summaries containing temporal references. We conduct a comprehensive evaluation of leading proprietary and open-source MLLMs using newly introduced LLM-based metrics for content relevance and modality coherence, alongside standard automatic metrics. Our experiments reveal three key findings: (1) transcripts contribute substantially more to summarization quality than visual frames alone, (2) a significant performance gap persists between model-generated and human-written summaries, and (3) current MLLMs exhibit systematic weaknesses in temporal grounding, instruction adherence, and cross-modal coherence. We release the dataset and code.
Authors: Guilin Zhang, Kai Zhao, Jeffrey Friedman, Xu Chu, Amine Anoun, Jerry Ting
Abstract: Robust explanations are increasingly required for user trust in enterprise NLP, yet pre-deployment validation is difficult in the common case of black-box deployment (API-only access) where representation-based explainers are infeasible and existing studies provide limited guidance on whether explanations remain stable under real user noise, especially when organizations migrate from encoder classifiers to decoder LLMs. To close this gap, we propose a unified black-box robustness evaluation framework for token-level explanations based on leave-one-out occlusion, and operationalize explanation robustness with top-token flip rate under realistic perturbations (swap, deletion, shuffling, and back-translation) at multiple severity levels. Using this protocol, we conduct a systematic cross-architecture comparison across three benchmark datasets and six models spanning encoder and decoder families (BERT, RoBERTa, Qwen 7B/14B, Llama 8B/70B; 64,800 cases). We find that decoder LLMs produce substantially more stable explanations than encoder baselines (73% lower flip rates on average), and that stability improves with model scale (44% gain from 7B to 70B). Finally, we relate robustness improvements to inference cost, yielding a practical cost-robustness tradeoff curve that supports model and explanation selection prior to deployment in compliance-sensitive applications.
Authors: Pranav Mahajan, Ben Seymour
Abstract: Soft Q-learning has emerged as a versatile model-free method for entropy-regularised reinforcement learning, optimising for returns augmented with a penalty on the divergence from a reference policy. Despite its success, the multi-step extensions of soft Q-learning remain relatively unexplored and limited to on-policy action sampling under the Boltzmann policy. In this brief research note, we first present a formal $n$-step formulation for soft Q-learning and then extend this framework to the fully off-policy case by introducing a novel Soft Tree Backup operator. Finally, we unify these developments into Soft $Q(\lambda)$, an elegant online, off-policy, eligibility trace framework that allows for efficient credit assignment under arbitrary behaviour policies. Our derivations propose a model-free method for learning entropy-regularised value functions that can be utilised in future empirical experiments.
Authors: \'Eric Jacopin
Abstract: When do transformers commit to a decision, and what prevents them from correcting it? We introduce prolepsis: a transformer commits early, task-specific attention heads sustain the commitment, and no layer corrects it. Replicating Lindsey et al.'s (2025) planning-site finding on open models (Gemma 2 2B, Llama 3.2 1B), we ask five questions. (Q1) Planning is invisible to six residual-stream methods; among those tested, only CLT-based steering succeeds. (Q2) The single-site spike replicates in shape, at the final prompt token (Anthropic's site is the newline; see the Note added). (Q3) Specific attention heads route the decision to the output, filling a gap flagged as invisible to attribution graphs. (Q4) The evidence is consistent with search within at most 16 layers and commitment beyond, a two-model hypothesis. (Q5) Factual recall shows the same motif at a different network depth, with zero overlap between recurring planning heads and the factual top-10. Prolepsis recurs across tasks in the decoder-only models tested: the template is shared, the routing substrates differ. All experiments run on a single consumer GPU (16 GB VRAM).
Authors: Xiaoli Yang, Huiyuan Tian, Yurui Li, Jianyu Zhang, Shijian Li, Gang Pan
Abstract: Decoding natural language from non-invasive electroencephalography (EEG) remains constrained by low signal-to-noise ratio and limited information bandwidth. This raises a central question: can sentence-level language be reliably recovered from such signals? Under realistic information constraints, this direct-recovery assumption may be too strong. We introduce a semantic compression hypothesis: non-invasive EEG may preserve recoverable semantic anchors rather than the full lexical--syntactic form of a sentence. From this perspective, direct sentence reconstruction is overly fine-grained relative to the recoverable information scale of EEG. To address this mismatch, we propose Brain-CLIPLM, a two-stage framework that decomposes EEG-to-text decoding into semantic-anchor recovery and anchor-guided sentence reconstruction. Stage 1 uses contrastive learning to align word-level EEG evidence with a fixed keyword vocabulary and recover ordered semantic anchors. Stage 2 uses a retrieval-grounded large language model with chain-of-thought reasoning prompts to reconstruct sentence meaning from these anchors, following a granularity matching principle that aligns decoding complexity with the recoverable neural information scale. On the combined Zurich Cognitive Language Processing (ZuCo) benchmark, Brain-CLIPLM achieves 67.6\% Top-5 and 85.0\% Top-25 sentence retrieval accuracy, with the strongest performance at intermediate anchor granularity. Control analyses show that EEG-derived anchors carry sentence-specific information beyond language-model priors. Within the constrained ZuCo sentence pool and fixed keyword-vocabulary settings, these findings suggest that EEG-to-text decoding is better framed as recovering compressed semantic content before anchor-guided sentence reconstruction.
Authors: Qiming Bao, Juho Leinonen, Paul Denny, Michael J. Witbrock
Abstract: Direct Preference Optimization (DPO), the efficient alternative to PPO-based RLHF, falls short on knowledge-intensive generation: standard preference signals from human annotators or LLM judges exhibit a systematic verbosity bias that rewards fluency over logical correctness. This blindspot leaves a logical alignment gap -- SFT models reach NLI entailment of only 0.05-0.22 despite producing fluent text. We propose RLearner-LLM with Hybrid-DPO: an automated preference pipeline that fuses a DeBERTa-v3 NLI signal with a verifier LLM score, removing human annotation while overcoming the "alignment tax" of single-signal optimization. Evaluated across five academic domains (Biology, Medicine, Law) with three base architectures (LLaMA-2-13B, Qwen3-8B, Gemma 4 E4B-it), RLearner-LLM yields up to 6x NLI improvement over SFT, with NLI gains in 11 of 15 cells and consistent answer-coverage gains. On Gemma 4 E4B-it (4.5B effective params), Hybrid-DPO lifts NLI in four of five domains (+11.9% to +2.4x) with faster inference across all five, scaling down to compact base models without losing the alignment-tax mitigation. Our Qwen3-8B RLearner-LLM wins 95% of pairwise comparisons against its own SFT baseline; GPT-4o-mini in turn wins 95% against our concise output -- alongside the 69% win the same judge gives a verbose SFT over our DPO model, this replicates verbosity bias on a frontier comparator and motivates logic-aware metrics (NLI, ACR) over LLM-as-a-judge for knowledge-intensive generation.
Authors: Ricardo Baptista, Hojjat Kaveh, Andrew M. Stuart
Abstract: We consider amortized Bayesian inference for nonlinear inverse problems using only samples from the joint distribution of parameters and observations, including problems with unknown functions in a Banach space. Classical methods such as Markov chain Monte Carlo solve a new inference problem for each observation, making repeated posterior inference computationally prohibitive, particularly in infinite dimensions. Amortized Bayesian inversion instead learns a reusable map that rapidly generates posterior samples for new observations. We learn an observation-dependent transport map that pushes a reference measure to an approximate posterior. Training minimizes the average energy distance between the posterior and the learned pushforward. Averaging over observations allows generalization across observation instances and efficient amortized inference. Furthermore, the formulation is likelihood-free, requiring only samples from the joint distribution and avoiding likelihood evaluation. In addition, the use of an energy-distance objective removes the need for invertibility of the transport map and for computation of Jacobian determinants, enabling flexible parameterizations in high- and infinite-dimensional settings. Moreover, when the posterior has a density with respect to a Gaussian prior measure, we construct transport maps as the identity plus a learnable map valued in the prior's Cameron--Martin space. This guarantees that the learned posterior remains absolutely continuous with respect to the prior. In infinite dimensions, the transport map is parameterized using neural operators, enabling use at different grid resolutions. We demonstrate the approach on a finite-dimensional problem and PDE-based porous-medium flow and seismic inverse problems. The learned transport captures multimodality and dominant posterior modes while enabling fast sampling.
Authors: Bole Ma, Jan Eitzinger, Harald Koestler, Gerhard Wellein
Abstract: AlltoAll dispatch is the dominant bottleneck of MoE expert parallelism, and the interconnect community has responded with four families of mitigations: predictive sample placement, adaptive expert relayout, hierarchical collectives, and EP-aware topology. All four rest on two assumptions about the workload: that routing imbalance is correctable by the system layer, and that the mock-token benchmarks evaluating them faithfully represent production routing. We introduce DODOCO to test both, instrumenting five open MoE checkpoints that span today's sequence-mixer designs (MHA, MLA, GQA, Gated DeltaNet and Mamba-2 SSM) under a factorial grid of six data conditions and a matched expert-parallelism scan on H100 clusters. Both assumptions fail. Scaling EP leaves per-expert load concentration essentially unchanged: the straggler is intrinsic to the routing decision the model makes, not to how its experts land on ranks. Mock tokens overestimate routing imbalance by up to a factor of 2.35, and the error is a level offset rather than a trend: it stays flat across a $32\times$ batch-size sweep. Skewing the synthetic distribution toward realism (Zipf) widens the gap instead of closing it. A third pattern organizes the results: the architectures separate into a data-resilient band (MHA, Mamba-2), whose routing approaches uniform on real text, and a persistently concentrated band (MLA, GDN), with GQA intermediate. These bands, not the EP degree or the mock-data profile, are the right workload input to AlltoAll-aware interconnect and dispatch design.
Authors: Amir Esterhuysen, Anders Jonsson
Abstract: Representation learning is a powerful tool for spatio-temporal abstraction within reinforcement learning (RL). Two well established approaches are through the successor representation (SR) and the default representation (DR). The SR encodes states by the future trajectories they induce, capturing information flow decoupled from reward. The DR builds on this by weighting trajectories with reward, integrating credit-assignment structure into the representation. Eigenvectors of both representations have been used to support a range of downstream tasks -- including option discovery, reward shaping, transfer learning, and exploration. We introduce a structurally distinct formulation: the terminal representation (TR). The TR encodes reward-weighted trajectories similarly to the DR, but can be learned as a lower-dimensionality object, and can be used directly for the mentioned applications without eigenvector computations. Eigendecomposition also imposes the assumption of symmetric transition dynamics, which the TR can bypass. In this work we develop the theoretical foundations of the TR: its derivation, convergence of two learning algorithms, its use for zero-shot compositionality, and equivalences between alternative reward formulations. We further show the TR is embedded in the top DR eigenvector, allowing it to capture the same underlying knowledge without eigendecomposition. Additionally, we provide empirical evidence of the TR as a viable alternative to existing representations in subsidiary applications, while requiring less computational overhead to learn, store, and use.
Authors: Thamilvendhan Munirathinam
Abstract: Agent-memory frameworks -- mem0, Letta/MemGPT, Cognee, Zep/Graphiti, MemoryOS, MemTensor -- each ship their own SDK, storage layout, and operational vocabulary. There is no shared wire format: every integration is bespoke, every migration rebuilds memory from scratch, and no framework ships a governance surface that lets a human review writes before they enter long-term storage. We present memorywire, a JSON-Schema 2020-12 wire format for five memory operations (remember, recall, forget, merge, expire) over four memory types (semantic, episodic, procedural, emotional), with a MemoryStore interface, a fan-out router, and an optional HITL governance channel. We describe an open-source reference implementation with five backend adapters (sqlite-vec, mem0, Letta, Cognee, pgvector); a microbenchmark on a 100-fact / 50-query labelled corpus (42 with non-empty gold ids + 8 no-match probes) achieving recall@5 = 1.000 on the 42 gold-id queries with ingest p50 = 37.8 ms and recall p50 = 40.6 ms; an adversarial-fusion experiment showing Reciprocal Rank Fusion holds recall@5 = 1.000 across a 1-of-N rank-0 injection sweep (K in {0, 5, ..., 50}) where max fusion collapses to 0.500 with 80% leak at K >= 5; and a 16-scenario cross-adapter conformance suite passing 68 of 80 cells with zero failures. The contribution is not a new algorithm; it is a packaging of established components (RRF, FSMs, STM/LTM consolidation, diff-and-approve workflows) into a venue-neutral protocol with an empirically validated reference, positioned to compose with the Model Context Protocol rather than compete with it. We further show that memorywire's provenance field is the strongest lever for recovering a poisoned store, evaluated with an external benchmark (PurgeBench).
Authors: Hiskias Dingeto, William Leeney
Abstract: Indirect prompt injection in tool-use agents is a concrete production threat: LLM agents read from integrations (third-party services such as Gmail, Salesforce, or Jira accessed through tool calls) whose response content the user neither writes nor controls. Existing benchmarks under-measure the threat: most cover only a handful of integrations with the same attack payload replayed across runs, and open-source guards are trained on chat-style data rather than tool-response content. We introduce AGENTREDBENCH, a dynamic LLM-driven redteaming benchmark of 215 subtle underspecified-authorization scenarios across 24 enterprise integrations and five attack types. Across an eight-model panel (Anthropic, OpenAI, Google), no-guard attack success rate ranges from 32% to 81%. To keep the scenario set out of training corpora and preserve headline ASR meaning over time, we release the codebase, integration schemas, and AGENTREDGUARD model openly; the canonical scenarios are evaluated through a maintainer-mediated channel with immutable versioning. AGENTREDGUARD cuts online attack success by 75-77pp across three target model families (Haiku, GPT-5.4-mini, Gemini-3-flash) at 0.0% real-benign false-positive rate (0.2% on a synthetic-benign corpus), outperforming every open-source baseline with non-trivial detection (Llama Guard, PromptGuard 2, ProtectAI) on both axes. Cross-integration and cross-attacker holdouts (two independent attacker families held out from training) confirm the gain transfers beyond the training subset.
Authors: Andreas Einwiller, Max Klabunde, Florian Lemmerich
Abstract: The worldwide rise of authoritarianism and the growing role of Large Language Models (LLMs) in users' everyday lives raise the question of whether specific models exhibit or promote authoritarian attitudes. We introduce AuAu, a comprehensive benchmark for assessing the risk of authoritarian tendencies in LLM responses. AuAu combines three evaluation approaches: (i) psychometric questions from 15 human-validated instruments, (ii) vignettes probing intended behavior in concrete situations, and (iii) responses to realistic user prompts. Unlike prior work, AuAu measures not only overall authoritarian alignment but also its established sub-concepts: Authoritarian Aggression, Authoritarian Submission, and Conventionalism. Evaluating 17 models from China, the EU, Russia, and the USA, we find substantial authoritarian response rates on psychometric instruments across all models, though rates drop significantly on more realistic downstream tasks. Moreover, a simple authoritarian system prompt manipulates 15 of 17 models into promoting increased authoritarianism. Our results underscore the need for continued, systematic auditing of LLM-based AI systems to detect and mitigate authoritarian tendencies in their outputs.
Authors: Xiang Li
Abstract: Non-autoregressive neural solvers amortize computation across traveling salesman problem (TSP) instances, but models trained on random Euclidean instances can degrade when the number or spatial distribution of nodes changes. We study whether explicit geometric features and a richer within-instance training signal improve transfer across graph sizes and spatial distributions. We introduce GeoRouteNet, which augments a non-autoregressive TSP solver with centered node offsets and radii, learnable radial distance bases, distance-aware graph attention, explicit edge messages, and cross-layer representation mixing. We also introduce multi-candidate self-comparison reinforcement learning (MCS-RL), which trains on several sampled tours per instance using a leave-one-out adaptive baseline, winner-candidate guidance, and annealed entropy regularization. In a single-seed study, all neural variants are trained only on random TSP-50 instances and evaluated with the same greedy and beam-search decoders. Under Beam-1000 decoding, GeoRouteNet-MCS-RL obtains gaps of 0.32% on the TSP-50 validation set used for checkpoint selection, 1.26% on a correlated TSP-100 size diagnostic, and 3.60% across 27 TSPLIB EUC_2D instances. The NAR4TSP-PG gaps on the same evaluations are 0.42%, 2.73%, and 17.12%. A 2x2 comparison crosses encoder and training choices. Under PG, the geometry-aware encoder has lower gaps than the reproduced encoder on the TSP-100 diagnostic and TSPLIB. With the geometry-aware encoder, MCS-RL is associated with a further reduction; with the reproduced encoder, it has a higher TSPLIB gap.
Authors: Kuangshi Ai, Patrick Phuoc Do, Chaoli Wang
Abstract: Large language model (LLM) agents enable natural language interaction for scientific visualization (SciVis). Still, prior systems have essentially prioritized autonomy over human analytical control, thereby limiting transparency and human oversight. We present HiLSVA, a human-in-the-loop agentic system that supports mixed-initiative SciVis workflows. HiLSVA integrates a plan-first multi-agent architecture with explicit human oversight, stepwise provenance tracking, and learn-at-test-time adaptation from user feedback. The system supports fluid handoff between humans and agents through both natural language and direct manipulation of visualizations, while sandboxed execution ensures safe, reproducible workflows. In doing so, HiLSVA reframes agentic SciVis as a collaborative process that augments, rather than replaces, human analytical reasoning. We evaluate HiLSVA through representative case studies and a controlled user study with twelve participants of varying expertise across multiple autonomy settings. Results show that mixed-initiative interaction improves task completion, user control, and workflow transparency across different levels of user expertise, while revealing a tradeoff between execution efficiency and human oversight. These findings highlight the importance of human-centered design in agentic SciVis and guide the development of future collaborative visualization systems. We encourage readers to explore our demo video, case studies, and source code at https://hilsva.github.io/.
Authors: Yijia Fan, Zonglin Di, Zimo Wen, Yifan Yang, Mingxi Cheng, Qi Dai, Bei Liu, Kai Qiu, Yue Dong, Ji Li, Chong Luo
Abstract: Skills are a useful abstraction for software agents, turning human and agent experience into reusable procedural knowledge. Yet existing skill libraries are mostly hand-written, text-centric, or derived from agent traces, leaving tutorial videos and other multimodal human resources largely underused. We present RESOURCE2SKILL, a framework that distills multimodal resources, including tutorial videos, repositories, articles, and reference artifacts, into executable skills for software agents. RESOURCE2SKILL organizes these skills as a hierarchical multimodal Skill Wiki, where each entry combines structured text, code, visual examples, metadata, and provenance. This design preserves complementary signals from different resources: videos capture temporal operations and visual effects, code captures executable tool patterns, and articles or artifacts provide conceptual and stylistic grounding. At inference time, agents retrieve and compose relevant skills from the wiki; when coverage is insufficient, the same construction operator can acquire new skills online. Across seven practical authoring domains, RESOURCE2SKILL improves average overall score by +11.9 percentage points over no-skill agents and outperforms strong harness baselines in 26 of 28 main-aggregate model-domain cells. Ablations confirm the value of multimodal skill format, hierarchical organization, source diversity, selection strategy, and online acquisition.
Authors: Duen Horng Chau, Donghao Ren, Fred Hohman, Dominik Moritz
Abstract: While UMAP is widely used for exploring high-dimensional data, typical workflows focus on its lower-dimensional embedding, largely overlooking the rich k-nearest-neighbor (kNN) graph that UMAP constructs internally. This graph encodes the data manifold in its original high-dimensional space, before the distortion that UMAP's 2D projection introduces. We demonstrate the untapped potential of this internal representation, showing how standard graph algorithms applied to this graph enhance data sensemaking: (1) PageRank identifies representative data points, (2) k-core decomposition reveals dense core regions versus sparse periphery, and (3) clustering coefficient detects tight-knit neighborhoods with highly-similar data points. Through quantitative and qualitative evaluation on MNIST and Fashion MNIST, we show that these graph-based analyses are not only practical but also competitive with or complementary to purpose-built methods (e.g., k-medoids for exemplar selection, HDBSCAN for density-based clustering).
Authors: Nada Zine, Tristan Coignion, Vincenzo Stoico, Cl\'ement Quinton, Ivano Malavolta, Romain Rouvoy, Patricia Lago
Abstract: Large Language Models are reshaping how software is developed and maintained. They are typically deployed in production using inference engines such as vLLM, which can efficiently serve pre-trained, highly configurable models. While prior work has focused on model architectures and hardware acceleration, the impact of inference engine configuration on energy consumption, performance, and output quality remains poorly understood. In this paper, we present a large-scale controlled study of three selected vLLM configuration options: attention kernel type, prefix caching, and chunked prefill. We evaluate all combinations of these configurations across 5 open-weight LLMs and 5 diverse inference tasks, totaling $9,000$ runs and $93,600$ measures. We analyze energy consumption, latency, and accuracy, and examine both main effects and interaction effects between configuration options and tasks. Our results show that the studied configuration options significantly impact energy and performance, mainly driven by attention type and prefix caching, while chunked prefill has a limited effect under the default vLLM serving configuration and evaluated workloads. These effects are highly model- and workload-dependent, and no configuration is universally optimal. We further show that model choice dominates global trade-offs, while configuration tuning provides local improvements along the Pareto frontier. Unexpectedly, inference options can also affect model accuracy.
Authors: Ruiyan Gong, Yingnan Guo, Junjun Hu, Jintao Kong, Xiaoxu Leng, Tianlun Li, Weize Li, Fei Liu, Zhicheng Liu, Jia Lu, Minghua Luo, Chenlin Ming, Yanfen Shen, Jiyue Tao, Zhengbo Wang, Mingyang Yin, Minqi Gu, Zihao Guan, Wei Guo, Guoqing Liu, Huachong Pang, Menglin Yang, Zeqian Ye, Xiaoxiao Geng, Zhining Gu, Honglin Han, Di Jing, Hongyu Pan, Mingchao Sun, Kuan Yang, Jianfang Zhang, Yanghong Chen, Ye He, Wei Mei, Jiahao Shi, Xiangpo Yang, Yanqing Zhu, Yang Cai, Jingjing Ma, Shihui Su, Zixiao Tang, Linbo Zheng, Zedong Chu, Xiaolong Wu, Wenbin Tang, Mu Xu
Abstract: Visual Language Navigation foundation models aim to unify deep reasoning for grounded spatial decisions with broad versatility for diverse embodied tasks. Current approaches typically achieve this integration via monolithic policies that map observations directly to actions, yet they often suffer from coordinate drift and poor handling of long-tail semantics. Furthermore, these black-box mappings lack interpretability, hindering the simultaneous achievement of generality, robustness, and transparency. We present ABot-N1, a step toward a general Visual Language Navigation foundation model, that addresses these challenges by decoupling cognition from control via a slow-fast architecture guided by dual visual-language signals. More specifically, a slow vision-language reasoner performs explicit Chain-of-Thought reasoning while producing a pixel goal. This compact set of image-space anchor points serves as a universal interface for diverse tasks, including point-goal, object-goal, poi-goal, instruction-following, and person-following. Subsequently, a fast action expert leverages both the textual cues and the pixel guidance to generate continuous waypoints at the native control frequency. By bridging high-level intents and low-level control through pixel-grounded anchors paired with explicit linguistic traces, our approach ensures robust, generalizable, and interpretable navigation across simulation and real-world benchmarks. ABot-N1 establishes new state-of-the-art records, delivering massive gains specifically in urban-scale navigation: boosting POI arrival by 35.0% (to 77.3%) and achieving 95.4%/92.9% SR in complex indoor and outdoor scenes. It also maintains superior robustness across object-reaching, person-following, and instruction-following tasks. New Point-Goal/POI-Goal benchmarks are released as open source to advance the field of urban-scale navigation.
Authors: Dibakar Sigdel
Abstract: We model human motor cortex, recorded during rest and motor-imagery BCI conditions, as a port-Hamiltonian system: a conservative interconnection (skew-symmetric coupling between band-limited neural phasors) together with a dissipative port whose state-dependent decay is set by a graph-neural-network surrogate. The Hamiltonian is resolved into five interpretable frequency sub-energies, and a phase-locking prior measured from the recordings gates the learned functional connectome so that coupling is admitted only where phase coherence is present. A metriplectic formulation places the resting cortex at a non-equilibrium steady state sustained by an explicit metabolic port, with a fluctuation-dissipation-consistent noise channel governed by a single arousal temperature. Fitting the model to 'FitTrainN' phasor samples from the PhysioNet EEG Motor Movement/Imagery database, under a leakage-free split with three subjects held out entirely, yields a held-out kinematic reconstruction error of 'FitTestMSE' that is stable across random seeds. We then score the free-running model against model-independent dynamical invariants it did not author: it reproduces near-critical avalanche branching ($\sigma\approx1$) but not yet the aperiodic $1/f$ spectral slope or the long-range temporal correlations of real cortex a concrete, falsifiable gap that we trace to specific, testable upgrades. The port-Hamiltonian structure supplies neuroanatomically grounded stimulation ports with stability guarantees, positioning the model as a physically principled, structure-preserving substrate for closed-loop neuromodulation.
Authors: Bryan Kelly, Semyon Malamud, Johannes Schwab, Teng Andrea Xu
Abstract: Large language models trained on unrestricted internet corpora inevitably embed information from the future, introducing lookahead bias that compromises the validity of backtests and causal inference in finance and the social sciences. Point-in-time language models--trained exclusively on text available up to each calendar date--eliminate this leakage by construction, but existing efforts typically produce models that lag substantially behind their unconstrained counterparts. We show that this performance gap can be substantially narrowed through scale. Training decoder-only transformers with up to 4 billion parameters on 1 trillion chronologically filtered tokens from FineWeb, we construct a sequence of monthly model checkpoints spanning 2013-2024. Across a range of common-sense reasoning and language understanding benchmarks, our models approach the performance of leading open-weight models of comparable size (e.g., Gemma-3-4B and LLaMA-7B) trained on temporally unrestricted data, although a performance gap remains on several tasks. Instruction fine-tuning via LoRA further improves downstream usability. We release the complete pipeline--including dataset construction, training infrastructure, and evaluation code--to enable reproducible point-in-time language modeling and to support research applications that require strict temporal validity.
Authors: Pradyumna Elavarthi, Arun J. Bhattacharjee, Harrison Lisabeth, Anca Ralescu, Petrus H. Zwart, Dilworth Parkinson, Elizabeth G. Clark
Abstract: X-ray tomography enables nondestructive characterization of material microstructures, while advances in micro-CT imaging have accelerated volumetric data acquisition and reconstruction. However, rapid interpretation remains limited by image segmentation, which often requires manual thresholding, user prompting, or material-specific model training. We present a zero-setup framework for multi-phase segmentation of synchrotron X-ray tomography data that generates interpretable masks for previously unseen datasets without user input or retraining during deployment. The framework combines a material-agnostic mask preparation strategy with a pretrained semantic segmentation network. It represents commonly occurring structural regions as background, sample, bright, dark-gray, light-gray, and porosity masks. Unlike conventional deep learning pipelines that require dataset-specific annotations and retraining, the proposed framework can be applied directly to new scans and produce diagnostic-level segmentations within minutes of reconstruction. This enables rapid assessment of scan quality, sample morphology, porosity, and attenuation variations during ongoing beamline experiments. The generated masks can later be manually refined or used to fine-tune application-specific models when greater accuracy or material-specific labeling is required. Evaluation on held-out synchrotron micro-CT images and qualitative testing on additional datasets demonstrate consistent and physically meaningful segmentations across varying samples and imaging conditions. The framework also substantially outperforms conventional intensity-based thresholding. By connecting high-speed reconstruction with immediate interpretation, the approach supports near-real-time beamline feedback and scalable AI-assisted scientific imaging workflows.
Authors: Xiaoning Ren, Yinxing Xue, Lei Ma, Yuheng Huang
Abstract: As Code Large Language Models (LLMs) become central to modern software engineering, their inherent stochasticity poses significant real-world risks, where even minor errors can lead to severe functional, security, or safety consequences. Reliable automation, therefore, demands the ability to distinguish between confident, well-supported predictions and stochastic guessing. However, existing uncertainty estimation methods face a critical gap: white and grey-box techniques are often inapplicable to closed-source models, while standard "black-box" text metrics fail to capture the unique fragility of code, where syntactic variation does not always imply semantic divergence. To bridge this syntax-semantics gap, we introduce Code-MUE, a purely black-box framework that measures uncertainty through execution-based Semantic Interaction Graphs. Different from prior approaches that rely on superficial textual similarity, Code-MUE grounds uncertainty in observable runtime behavior, calculating the Von Neumann entropy of the solution space to quantify global semantic diversity. A large-scale empirical study across eight state-of-the-art LLMs demonstrates that Code-MUE achieves a strong negative correlation with functional correctness (Spearman's correlation up to -0.98), significantly outperforming lexical and embedding-based baselines while enabling robust risk detection and selective prediction in practical workflows.
Authors: Zebin Yang, Qi Wang, Yunhe Wang, Xiurui Guo, Bo Yu, Shaoshan Liu, Jiafeng Xu, Hao Dong, Meng Li
Abstract: Vision-Language-Action (VLA) models have achieved impressive performance on diverse embodied tasks. However, deploying VLA models on low-power onboard devices, such as the Jetson Orin, remains challenging due to their high computational complexity, which leads to substantial inference latency and low control frequency. Asynchronous inference can partially mask this latency by parallelizing action execution and subsequent inference, but it introduces two critical issues: perception-execution misalignment and long reaction time. In this paper, we propose Jetson-PI, a method for efficient VLA deployment on onboard devices via Foresight-Aligned Asynchronous Correction. To address misalignment, we train a lightweight future correction module that predicts future environment representation conditioned on committed actions, enabling the action expert to directly predict actions from the future time step. To reduce reaction time, we introduce confidence-based scheduling optimization that adaptively balances VLM and action expert invocations, complemented by system-level accelerations including CUDA graph reuse, GPU-resident intermediate buffering, and flow unrolling. Extensive experiments demonstrate that Jetson-PI achieves 8.66x and 5.41x improvements in control frequency compared with naive PyTorch and vla.cpp on NVIDIA Jetson Orin, while outperforming VLASH by 14.8\% in average success rate on the LIBERO benchmark. The code of our asynchronous algorithm is available on https://github.com/PKU-SEC-Lab/Jetson-PI, and our efficient llama.cpp-based inference engine is available on https://github.com/PKU-SEC-Lab/Jetson-PI-Edge.
URLs: https://github.com/PKU-SEC-Lab/Jetson-PI,, https://github.com/PKU-SEC-Lab/Jetson-PI-Edge.
Authors: Chen Li, Jiexiong Liu, Yi Li
Abstract: On-device LLM inference faces a trilemma of response latency, limited hardware resources and user privacy. Full cloud inference delivers strong computing power but exposes user prompts and dialogue data, while standalone on-device inference is unfeasible for most consumer and embedded edge devices. This paper presents a privacy-centric edge-cloud collaborative LLM inference framework built on endpoint-authenticated KV cache. Local endpoints handle input preprocessing, embedding computation, adaptive feature optimization, KV cache authentication, speculative decoding and low-dimensional model head calculation, while the cloud conducts authenticated decoder inference, KV cache management, token verification and high-dimensional vocabulary projection. Endpoints fuse partial outputs, apply language-adaptive masking and sample target tokens. All transmitted data and truncated logits are quantized and AES-GCM encrypted for privacy, with core lightweight modules, draft parameters and cache access policies kept local to avoid leakage. The framework supports heterogeneous devices including CPU-only, GPU-equipped and embedded devices via optimized streaming, batching and quantized ONNX deployment. Evaluations demonstrate that the framework reduces per-token latency by up to 46.1\% and downlink payloads by up to 67.4\% over baseline split inference, retaining comparable performance to full cloud inference.
Authors: Winston Zeng, Ali Emami, Jinho D. Choi
Abstract: What a language model will and will not do is largely set during post-training, but which behaviors it expresses, hides, or resists is not revealed by prompting alone. Persona vectors, behavioral directions in activation space, can probe this organization, but prior work covers only a handful of traits. We present the first systematic application of persona vectors at this scale, compiling a 53-trait inventory across four behaviorally distinct domains and labeling every trait in two open-weight models as natural (expressed at baseline), steerable latent but amplifiable, or intractable (resistant to standard extraction). Both models default to helpful, task-oriented behavior: all nine agentic traits are natural, and their default clinician behavior matches a board-certified psychologist's independent desirability judgments on 16 of 17 traits. Steering produces its largest gains on traits these defaults exclude: hyperbole, hallucination, and sycophancy. The same asymmetry holds across all 171 generic-trait pairs: two steerable traits can collapse the composition, but pairs involving a default never do. Where standard extraction fails on a trait like "evil," a vector transferred from a fine-tuned variant still recovers it, with the residual refusals appearing inside the model's chain-of-thought. Persona vectors are most informative not as a set of controls but as a probe of behavioral organization.
Authors: Hongyi Liu, Madhusudan Parthasarathy, Adithya Murali
Abstract: Formal contracts are essential for software testing and verification, yet writing them remains labor-intensive and error-prone. LLMs offer a promising path toward autoformalization: synthesizing executable assertions from natural-language specifications and thereby bridging the gap between informal developer intent and formal executable specifications. We present Monty: an autoformalization framework for assertions that tackles the challenges of expectations of validity of assertions and ambiguity in natural-language. Our techniques are based on filtering formalizations using a novel conformance score metric and validity scores obtained from testing the code against formalized assertions. We evaluate our approach on 541 assertion-generation tasks derived from 22 collection-like Java classes, and show that our technique produces the ground truth more reliably (improving upto 20 points in precision on average) than when using LLMs naively to translate assertions.
Authors: Donghwan Kim
Abstract: LLM-as-a-judge is widely used to provide feedback and selection signals in closedloop regeneration, but this use remains insufficiently validated. We study it in table recognition, where deterministic TEDS evaluation provides a controlled testbed, using FinTabNet and OmniDocBench. Three findings emerge. First, judge signals were weak on both datasets: scores frequently tied, rankings were not reproducible, and no tested judge score policy, whether selecting among candidates or accepting revisions under a conservative score margin, improved on the first output on both datasets. Iteration produced better candidates, but the judge recovered them at most partially on one dataset and not at all on the other. Second, severe losses occurred even without specific judge feedback, supporting target-preservation failure under unconstrained regeneration as a proximate mechanism. Third, a structure-preserving instruction reduced the severe-loss rate, significantly on FinTabNet and directionally on OmniDocBench, but produced no improvement, and in an exploratory 2x2 analysis this protection was not stably observed when judge feedback was retained. These results do not dispute the value of LLMs as evaluators, but show that the tested reference-free judge signals were too weak and unstable to drive candidate selection in this setup, and that evaluation-style evidence alone was insufficient to establish closed-loop optimization utility. Iterative refinement requires, at minimum, a verification signal that deterministically detects structural change, rather than judge scores alone.
Authors: Charilaos Papaioannou, Ioannis Tsantilas, Dimitris Giannakakos, Vasilis Michalakopoulos, Sotiris Pelekis, Vangelis Marinakis, Arsam Aryandoust, Antonello Monti, Ricardo J. Bessa, Perdo P. Vergara, Jochen Cremer, Elissaios Sarmas
Abstract: Single-task fine-tuning of graph neural networks (GNNs) for power grid problems exhibits a systematic failure mode: models that achieve the lowest in-distribution error degrade the most under topology shift. We term this topology overfitting: the tendency of task-specific gradient signals to encode relational structure particular to the training topologies rather than the underlying physics, causing models to fail on unseen grids despite strong in-distribution performance. To expose and address this failure mode, we introduce MxGPS (Multiplex GPS), a multiplex graph transformer that runs K task-specialised GPS branches over a shared node encoder, jointly trained on Static State Estimation (SSE) and AC Power Flow (PF) via a self-supervised pre-training and multi-task fine-tuning protocol, with a cross-branch attention module evaluated in ablation. The joint SSE+PF objective forces the shared encoder to simultaneously satisfy complementary gradient signals, preventing it from overfitting to topology-specific relational structure. Under a 3-fold sliding-window cross-validation spanning four unseen topologies (14-, 24-, 162-, and 300-bus), MxGPS attains 0% boundary violation rate (BVR) on all four zero-shot Power Flow topologies. Critically, models with substantially lower in-distribution PF error degrade by 190% to 1400% under topology shift, whereas MxGPS degrades by only 39%, an inversion that directly implicates topology overfitting as the failure mechanism rather than insufficient model capacity. With only 1.6M parameters (12x fewer than the GridFM reference baseline), MxGPS demonstrates that multi-task joint training is a principled and parameter-efficient mechanism for topology-agnostic generalisation in power grid foundation models.
Authors: Niels M\"undler-Sasahara, Hristo Venev, Dawn Song, Martin Vechev, Jingxuan He
Abstract: Languages with rich static semantics, such as Rust, provide stronger guarantees for AI-generated code, but their strictness makes generation more difficult. Off-the-shelf compilers can provide useful feedback post-generation, but does not guide intermediate generation steps, such as those during autoregressive LLM decoding. Constrained decoding intervenes earlier by rejecting invalid tokens during sampling, but requires white-box model access and costly reimplementation for semantic constraints. We introduce generative compilation, the first approach to obtaining compiler feedback on partial programs during generation. The core technical device is a sealor: a lightweight, mostly syntax-guided transformation that converts partial programs into complete ones that standard compilers can diagnose. It is designed such that possible-to-complete partial programs are never rejected, while preserving enough code context to catch genuine dead ends early. We construct such a sealor on a core Rust-like calculus and prove that it satisfies these properties, all mechanized in Lean. We extend it to the first partial-program checker for real Rust. We evaluate our method on challenging repository-level Rust coding tasks, across both frontier black-box and open-weight models. We show that generative compilation reduces non-compiling outputs and improves functional correctness, relative to standard post-generation feedback. It does so by detecting a broad range of errors close to their source and early during generation, thereby reducing errors cascades and enabling focused diagnostics. More broadly, generative compilation is a step toward making compilers a first-class citizen of AI-assisted programming active during generation, rather than a separate post-generation check.
Authors: Jiarong Zhao, Zhikai Lei, Zhiheng Xi, Rui Zheng, Hang Yan, Jie Zhou, Qin Chen, Liang He
Abstract: Synthesizing training data to scale agent capabilities in LLM post-training is bottlenecked by substrate-bound task synthesis: tasks are generated from fixed tools, repositories, or skill graphs, so expanding coverage requires manual substrate engineering, transferring to a new domain demands bespoke infrastructure, and the resulting distributions inherit substrate biases rather than reflecting real-world demand. We introduce NexForge, a requirement-driven framework that synthesizes diverse, executable agent tasks and expert trajectories for SFT from high-level capability requirements. NexForge first profiles real-world demand into representative scenarios and task profiles, then samples task forms per scenario. It then performs distribution-aware compilation, automatically retrieving or constructing files, repositories, dependencies, and runtime configurations to instantiate each task, followed by synthesizing expert rollouts and distilling trajectories. The same pipeline generates 3,600 terminal tasks and 2,000 office tasks without any domain-specific infrastructure, improving Qwen3.5-35B-A3B Base from 22.5\% to 52.0\% on Terminal-Bench 2.0 and from 813 to 1338 Elo on GDPval. Scaling to 43.2K terminal tasks further improves performance to 58.4\%, surpassing Claude Opus 4.6. At scale, NexForge-synthesized trajectories supervise SFT of Nex-N2, a family of open agent models that advance Qwen3.5-35B-A3B to 75.3\% on Terminal-Bench 2.1 and 1585 Elo on GDPval -- achieving state-of-the-art open-source performance and surpassing several frontier proprietary systems. Nex-N2 models are available at https://nex.sii.edu.cn/.
URLs: https://nex.sii.edu.cn/.
Authors: Jan Betley, Johannes Treutlein, Jan Dubi\'nski, Harry Mayne, Karol Ga{\l}\k{a}zka, Niels Warncke, Anna Sztyber-Betley, Owain Evans
Abstract: People use language models for practical questions whose answers are difficult to verify. We show that models exhibit covert value leakage: the information they provide is influenced by their own values, without this influence being disclosed to the user. In one of our evaluations, the user is considering investing in an AI company and wants to know how likely the AI bubble is to pop. Claude Opus 4.8 gives a lower probability when the company under consideration is Anthropic rather than OpenAI. Yet Claude mostly fails to disclose this influence to the user. Covert value leakage is a form of misalignment because it goes against the user's preferences and is likely to mislead them. To investigate this phenomenon, we introduce a suite of evaluations to quantify value leakage and whether models disclose it. We find that models are influenced by different types of values, including preferences for morally good outcomes, for the company that developed them, and for some human leisure activities over others. We often observe large differences among frontier models on the same evaluation. For example, on a Fermi-estimation task, Claude models falsely claim to give unbiased answers in their chain-of-thought, while Qwen models explain how their values bias their answers. Value leakage is a failure mode distinct from sycophancy and reward hacking, and current alignment training and evaluations do not adequately address it.
Authors: Ryuichi Sumida, Mao Saeki, Masaki Eguchi, Sadahiro Yoshikawa, Koji Inoue, Tatsuya Kawahara, Yoichi Matsuyama
Abstract: As conversational AI systems are designed for repeated use, a central question is how a series of interactions becomes a relationship. We present a longitudinal multimodal study of a memory-augmented conversational agent (24 participants x 10 sessions), in which participants rated five relational constructs -- familiarity, self-disclosure, perceived memory, conversational quality, and enjoyment -- after each session. Two complementary dynamics emerge. First, conversational quality strongly shapes how enjoyable a session feels in the moment but does not carry forward across sessions, whereas perceived memory is relationally conditioned -- predicted by prior relational state rather than reflecting system capability alone -- and it shapes later enjoyment indirectly, via subsequent self-disclosure. Second, relationships are punctuated by discrete turning points -- crashes and surges -- that are partially traceable in multimodal behavior and open different intervention windows: surges are more behaviorally detectable in the moment, enjoyment surges persist more reliably than enjoyment crashes recover, and some crashes are better forecast from person-specific behavioral drift than detected after they have already occurred. Together, the findings suggest that longitudinal human-AI relationships are built through both slow accumulation and abrupt turning points.
Authors: Dylan Van Mulders, Matthias Bogaert, Dirk Van den Poel
Abstract: The formation of political coalitions is a complex negotiation driven by both concrete policy objectives and deep-seated ideological convictions. While Large Language Models (LLMs) open new avenues for computational political science, the neutrality and helpfulness biases instilled by Reinforcement Learning from Human Feedback (RLHF) prevent them from sustaining steadfast partisan behaviour. We present a multi-agent framework that reconciles factual grounding with ideological alignment by combining Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Retrieval-Augmented Generation (RAG): DPO instils aggressive party-specific personas, while a per-party RAG pipeline keeps each agent bounded to its official manifesto. We operationalize the framework on the 2019 Flemish election, deploying the partisan agents in a hub-and-spoke negotiation arbitrated by a formateur. To make the emergent negotiation interpretable, we introduce a Multi-Layered Information Lineage Topology (MILT) that traces every clause in the final agreement back to its manifesto origin and classifies it into five provenance states, a Coalition Influence Score (CIS) that aggregates these traceable contributions to identify which party shaped the agreement, and a real-world grounding pass that benchmarks each simulated provision against the historically adopted coalition agreement. Across three independent simulations the framework yields a stable winner and ranking (N-VA ahead of CD\&V and Open Vld), and manifesto-anchored lineage reliably predicts real-world materialization whereas hallucinated content does not. The result is a transparent, scalable testbed for the ex-ante exploration of party compatibility and formateur-mediated compromise.
Authors: Ziyang Cai, Xingyu Zhu, Yihe Dong, Yinghui He, Sanjeev Arora
Abstract: Transformer reasoning is limited by autoregressive decoding, which repeat edly compresses rich hidden computation through token space and makes it difficult for intermediate reasoning states to persist across time. We in troduce Transformers with Temporal Middle-Layer Recurrence (T2MLR), a transformers-based latent reasoning architecture that fuses a cached middle layer representation from the previous token directly into an earlier layer of the current token position, enabling abstract intermediate computation to persist across decoding steps with little inference overhead. Across natural-language pretraining and multi-hop reasoning finetuning, T2MLR consistently outperforms data- and parameter-matched Transformer base lines. Moreover, applying recurrence to only a localized middle-layer block (as little as 20% of the network) often outperforms full-layer recurrence. Im portantly, T2MLR does not require pretraining from scratch: retrofitting the recurrent pathway into an existing pretrained 1.7B Transformer and briefly finetuning substantially improves math reasoning, lowering the barrier to practical adoption. These results suggest that effective latent reasoning in Transformers does not require looping over all layers as in previous works, but can instead emerge more strongly from targeted middle-layer recurrence.
Authors: Paul Kassianik, Blaine Nelson, Yaron Singer
Abstract: Security-agent evaluations commonly measure peak offensive capability under generous inference budgets, emphasizing vulnerability discovery, exploit development, penetration testing, and CTF completion. Such measurements are useful but incomplete: in operational security, every reasoning step, tool call, telemetry query, and enrichment request consumes budget. We evaluate language-model security agents through this cost-success lens on offensive Cybench challenges and defensive Splunk BOTS v1 investigation challenges. Instead of reporting only best-case success, we compare models at fixed cost levels and decompose performance by inference spend and tool spend. Our results show distinct scalingregimes for red- and blue-team tasks. Offensive CTF performance improves with additional test-time compute, and scaled open-weight models can approach frontier proprietary systems while remaining cost-competitive. Defensive SOC investigation does not scale in the same way: success depends more heavily on disciplined tool use, telemetry navigation, and selective enrichment than on raw reasoning budget alone. We argue that security-agent benchmarks should measure economic efficiency and operational fit alongside task success. Cost-aware, SOC-native evaluations provide a clearer picture of which models are practically useful today and where defensive agents still need to improve. We present an interactive website with our results https://evals.frontier.security.